Cover generation method, apparatus, device, and medium

Abstract

This application discloses a cover generation method, apparatus, device, and medium, applied in the field of Internet technology. The method includes: displaying a content publishing page for publishing a target video; responding to a trigger operation on the cover generation entry point, displaying a target cover generated according to a cover generation strategy related to the target video's target category; the content publishing page is used to publish target content containing the target video and whose cover is the target cover; wherein the target cover is obtained by image processing of target video frames in the target video using the cover generation strategy, the target video frame being a video frame matching the cover generation strategy, and the cover generation strategy being summarized by analyzing the cover covers of related published content belonging to the target category, wherein the click-through rate of the content covers of the related published content is higher than a preset click-through rate threshold. Using this application embodiment can improve the quality of cover generation.

Description

Technical Field

[0001] This application relates to the field of Internet technology, specifically to the field of data display technology, and in particular to a cover generation method, apparatus, device, and medium. Background Technology

[0002] In today's digital age, video has become a widely used and shared media format. Therefore, people can publish content on content platforms. In this context, a high-quality cover image is crucial for attracting other users to view the video and increasing click-through rates. Traditional methods for determining cover images typically involve extracting keyframes or using the first frame as the cover. However, such methods are relatively simplistic, often resulting in cover images that lack appeal and representativeness, are of low quality, and fail to meet the needs of the video content and style, leading to poor click-through rates. Summary of the Invention

[0003] This application provides a cover generation method, apparatus, device, and medium, which can improve the quality of cover generation through a more intelligent cover determination method, thereby improving the click-through conversion effect for videos.

[0004] On one hand, embodiments of this application provide a cover generation method, the method comprising: Displays the content publishing page for publishing the target video; the content publishing page includes a cover image generation entry point; In response to a trigger operation on the cover generation entry, the target cover generated according to the cover generation strategy related to the target video's target category is displayed; the content publishing page is used to publish target content that includes the target video and whose content cover is the target cover; The target cover is obtained by image processing of the target video frame in the target video through a cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of related published content is higher than the preset click-through rate threshold.

[0005] On one hand, embodiments of this application provide a cover generation apparatus, the apparatus comprising: The content publishing module displays the content publishing page for publishing target videos; the content publishing page includes a cover generation entry point. The cover generation module is used to respond to the trigger operation of the cover generation entry and display the target cover generated according to the cover generation strategy related to the target video's target category; the content publishing page is used to publish target content that contains the target video and whose content cover is the target cover; The target cover is obtained by image processing of the target video frame in the target video through a cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of related published content is higher than the preset click-through rate threshold.

[0006] On one hand, embodiments of this application provide an electronic device including a processor and a memory, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute some or all of the steps in the above method.

[0007] On one hand, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform some or all of the steps in the above-described method.

[0008] Accordingly, according to one aspect of this application, a computer program product or computer program is provided, which includes computer instructions that, when executed by a processor, can implement some or all of the steps in the above-described method.

[0009] In this embodiment, a content publishing page for publishing the target video can be displayed, and a cover generation entry can be shown on the content publishing page. By triggering the cover generation entry, a target cover for the target video can be intelligently generated, facilitating the publication of target content containing the target video and whose cover is the target cover. The generation of the target cover can be achieved by processing the target video frames in the target video according to a cover generation strategy related to the target video's target category. A target video frame refers to a video frame that matches the cover generation strategy. In other words, the cover generation strategy related to the target video's category can be used to perform intelligent image processing on video frames that semantically match the cover generation strategy, and the processed target video frame can be used as the intelligently generated target cover. This cover generation strategy involves processing the cover of the content published in the relevant category. The analysis of cover images reveals that the click-through rate (CTR) of related published content covers exceeds a preset CTR threshold. This means that we can analyze a large number of high-quality content covers (i.e., content covers of published content with high CTR) within the target category in advance to obtain patterns in content covers that attract online users within that target category (such as image structure, filter style, text layout, etc.). This leads to a corresponding cover generation strategy. By using the cover generation strategy for intelligent cover generation, we can obtain high-quality covers that conform to popular preferences and fit the video content. This allows for a more intelligent cover determination method to improve the quality of cover generation, making the determined covers attractive and representative, meeting the needs of the video content and style, and thus improving the video's click-through conversion rate (i.e., the probability of users clicking on the content cover to view the published content). Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of a network architecture provided for an embodiment of this application; Figure 2 A schematic diagram illustrating a cover generation scenario provided in an embodiment of this application; Figure 3 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 1 ; Figure 4 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 2 ; Figures 5a-5e This application provides a schematic diagram of a cover selection scenario. Figure 6 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 3 ; Figures 7a-7c This application provides a schematic diagram of a cover editing scenario. Figure 8 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 4 ; Figure 9 This is a schematic diagram illustrating a smart cover generation scenario provided in an embodiment of this application; Figure 10 A flowchart illustrating a cover generation method provided in this application embodiment is shown in Figure 5. Figure 11 This is a schematic diagram of the structure of a cover generation device provided in an embodiment of this application; Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] The cover generation method proposed in this application is implemented in an electronic device, which can be a server or a terminal. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet computer, laptop computer, desktop computer, etc., but is not limited to these.

[0013] In the description of this application's embodiments, the published content can refer to user-posted notes (including text and images), short videos, mid-length videos, etc., on a content platform. Alternatively, it can also include live streams, instantaneous content (a type of short content that is shared and exists in real time), reviews of products or any type of interest (such as locations, music, etc.) or media data (such as movies, music), comments in notes, bullet comments, topic discussions, interest points (such as group chats, live streams, etc.), routes (such as cycling routes, travel guides, etc.). The specific content information included in the published content can be adjusted accordingly for different scenarios, and the content information in the published content can be determined by the publisher (such as the target account). No specific type of published content is limited here. The published content is stored on a server and distributed to users browsing the content platform through a content distribution algorithm deployed on the server. For example, when a user enters the platform's homepage, published content can be recommended to the user according to the content distribution algorithm.

[0014] The published content can be divided into video content (including video, title, and body text) and text / image content (including images, title, and body text). For video content, users primarily extract information from the video itself; they determine the content's message by watching the video. For text / image content, users primarily extract information from the text (and images); they determine the content's message by viewing the text (and images). When recommending published content to platform users, the first thing displayed is the content's cover image. For example, the platform's homepage or search results page displays a content card with the cover image. This cover image can then attract users to click on the content card to access the content details page, where they can view the video or images.

[0015] Therefore, the cover image is a key factor in attracting users to click and view published content. This application aims to utilize cover generation strategies related to business categories to generate high-quality covers that better match user preferences, thereby attracting users to the content details page to view relevant information and improving the interaction conversion of published content (such as using cover generation strategies related to the target category of the target video to target the cover image, so that when users view the target cover image, they can click on the target cover image or the content card it belongs to to enter the content details page to view the target video, thereby improving the interaction conversion of the video, such as increasing the click-through rate of the published content or the video).

[0016] The cover generation strategy refers to one or more strategies summarized by analyzing the cover images of related published content under the target category. This strategy can be used to characterize the user's click motivation for the cover, the cover's image design logic, and the cover's text design logic. At this point, the image processing method can be determined based on the cover generation strategy (such as how the image should be designed, how it should be processed, and what text should be added). When generating image processing prompts based on the image processing method, the image processing prompts are input into the image generation model (such as a text-to-image model, which generates an image by inputting text prompts, or an image-to-image model, which generates a new image based on a reference image by inputting a reference image and text prompts). The image generation model then generates the corresponding target cover. This target cover can contain the information in the content that the user wants to publish and also conform to the cover generation strategy. Since this cover generation strategy is based on the analysis of online content covers, the target cover conforms to the preferences accumulated by online users, thereby ensuring the quality of the target cover.

[0017] For example, a cover strategy library can be pre-built to store cover generation strategies related to each category in the category set. For instance, a category set can be constructed (e.g., a category set containing multiple categories based on the categories to which published content belongs on the content platform). In this case, the associated published content for each category can be obtained, and cover analysis can be performed on the cover images of the associated published content for each category to obtain the cover generation strategy related to each category. A category can be associated with one or more cover generation strategies.

[0018] In this context, "related content" can refer to content with a click-through rate (CTR) exceeding a preset CTR threshold. Since content attracts users to click and view through its cover image, cover images of content with high CTR can be considered high-quality. That is, high-CTR content within a specific category can be acquired as related content, and the cover images of these related content can be analyzed to obtain one or more cover generation strategies. The specific methods for generating cover images and target covers can be found in the descriptions of the following embodiments.

[0019] This application allows users to input target content to be published on a content publishing page, and then generate a target cover based on that content. The target content can be video, text (such as title, body text, etc.), or images. For example, if the content publishing page contains uploaded videos, the target cover can be generated based on the video content (e.g., using a cover generation strategy for the video's category to perform image processing on a specific video frame to obtain the target cover). Similarly, if the content publishing page contains text, the target cover can be generated based on the text content (e.g., using a cover generation strategy for the text's category to generate an image to obtain the target cover). Furthermore, if the content publishing page contains both text and images, the target cover can be generated based on both text and image content (e.g., using a cover generation strategy for the text's category to perform image processing on a specific image to obtain the target cover). No further limitations are imposed.

[0020] The term "in response to" indicates a state where a corresponding event occurs or a condition is met. The timing of subsequent actions performed in response to this event or condition is not necessarily strongly correlated with the time when the event occurs or the condition is met. For example, "in response to XX operation, execute the target step" means that the target step is triggered upon detection of "XX operation".

[0021] For example, in some cases, subsequent actions can be performed immediately when the event occurs or the condition is met; while in other cases, subsequent actions can be performed some time after the event occurs or the condition is met, meaning there can be an intermediate process between responding to the corresponding event or condition and the subsequent actions performed.

[0022] The term "trigger operation" refers to an action performed on certain information (such as a control). This operation sends a corresponding instruction to the electronic device, triggering it to execute the next task. The next task triggered by different trigger operations on different information can all be pre-set. This trigger operation can be a contact-based operation such as clicking, double-clicking, long-pressing, or swiping, or other preset event operations, or it can be a non-contact operation (such as voice instructions). In some cases, the trigger operation can also be executed by the electronic device based on a pre-defined program. No further limitations are specified here.

[0023] One of the network architecture diagrams proposed in the cover generation method is as follows: Figure 1 As shown, the network architecture may include server 100 (the number of servers is not limited) and a cluster of terminal devices (the number of terminal devices is not limited, such as terminal device 200a, terminal device 200b, ..., terminal device 200n), wherein communication connections may exist between servers. Simultaneously, a server may have a communication connection with any terminal device, so that the server can interact with the terminal device through this communication connection. The aforementioned communication connection is not limited in method; it can be a direct or indirect connection via wired communication, a direct or indirect connection via wireless communication, or other methods, which are not limited herein. Furthermore, it is understood that the electronic devices involved in the embodiments of this application may be... Figure 1 The terminal device shown can also be Figure 1 The server shown.

[0024] It should be understood that, such as Figure 1 Each terminal device in the shown terminal device cluster can have an application client installed for cover generation. This application client can be of any type, such as a social networking client, instant messaging client (e.g., a conferencing client), entertainment client (e.g., a live streaming client), multimedia client (e.g., a video client), information client (e.g., a news client), shopping client, or any other client capable of displaying text, images, audio, and video data. No specific type of application client is limited here.

[0025] For example, an application client refers to a client that can send and receive internet messages in real time and has information functions. Specifically, terminal devices (such as…) Figure 1The application client on the terminal device 200a) shown in the figure displays a content publishing page, which displays the uploaded target video. The content publishing page also displays a cover generation entry. When the cover generation entry is triggered, the target video can be sent to the server so that the server can generate a target cover for the target video and display it on the content publishing page by the application client.

[0026] Optionally, the aforementioned terminal devices and servers can be logically separated. Therefore, when referring to terminal devices and servers below, they may be physically the same device or different devices.

[0027] For further information, please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of a cover generation scenario provided in an embodiment of this application. The content publishing page 20a can display the uploaded target video to be published, and the content publishing page can display a content cover 20b (e.g., the first frame of the target video by default). Users can also enter text information (such as title, body text, etc.) on the content publishing page. This text information can be manually entered by the user or intelligently generated (e.g., intelligently generating a title based on the video content of the target video).

[0028] The content publishing page can display a cover generation entry 20c. After triggering the cover generation entry, the target cover 20d intelligently generated for the target video can be displayed on the content publishing page (if an intelligent cover is generated, the user can choose whether to regenerate it).

[0029] Users can choose whether to use the target cover on the content publishing page. After choosing to use the target cover, the content cover displayed on the content publishing page will be updated to the target cover.

[0030] Among them, performing the content publishing operation on the content publishing page allows you to publish target content that includes the target video and whose cover image is the target video.

[0031] The target cover is generated according to the cover generation strategy related to the target video's target category. For example, it is obtained by image processing of the target video frame in the target video through the cover generation strategy. The target video frame is the video frame in the target video that matches the cover generation strategy. This can enrich the way the cover is determined and make the cover confirmation more intelligent. At the same time, using the cover generation strategy accumulated for the target category to generate the target cover can make the target cover more in line with public preferences, thereby improving the quality of the target cover. After publication, it can attract more users to click to enter the target published content to view the target video and improve interaction and conversion. In specific embodiments of this application, scenarios involving the acquisition of user information and related data, such as the acquisition of user information (e.g., published content), require user permission or consent. That is, when these embodiments are applied to specific products or technologies, the collection, use, and processing of relevant user data comply with the relevant laws, regulations, and standards of the relevant regions. For example, interactive pages can be used to provide prompts indicating which data will be collected or acquired. Specifically, lists or other methods can be used to present the types and content of this data to the user. Data collection and processing will only proceed after a confirmation or instruction to allow data collection is received on the interactive page.

[0032] The scenarios described above are merely examples and do not constitute a limitation on the application scenarios of the technical solutions provided in this application. The technical solutions of this application can also be applied to other scenarios. For example, as those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0033] Based on the above description, this application proposes a cover generation method, which can be executed by the aforementioned electronic device, specifically... Figure 1 The terminal device shown. Please refer to [link / reference]. Figure 3 , Figure 3 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 1 .like Figure 3 As shown, the cover generation method of this application embodiment may include the following steps: S101. Display the content publishing page used to publish the target content.

[0034] The target content can be a target video, target text (such as a title and / or body text), target image, etc. The content publishing page can include a cover generation entry point. This cover generation entry point is used to intelligently generate a target cover for the target content.

[0035] For example, when the content publishing page displays the uploaded target video, the publishing content published through the content publishing page is video publishing content. The main subject of the video publishing content is the video. Therefore, the target content used to generate the target cover can be the target video. That is, the corresponding target cover is generated based on the video content of the target video to attract users.

[0036] For example, when the content publishing page displays the target text entered by the user, the published content through the content publishing page is image and text content. The main subject of information extraction in image and text content is text, so the target content used to generate the target cover can be the target text.

[0037] For example, when the content publishing page displays the target text and target image entered by the user, the published content published through the content publishing page is graphic content. The main information extraction subjects of graphic content are text and image. Therefore, the target content used to generate the target cover can be the target text and the target image.

[0038] The content type of the target content is not limited here, and the specific method for generating the target cover can be found in the relevant description of the following embodiments.

[0039] S102. In response to the trigger operation for the cover generation entry, display the target cover generated according to the cover generation strategy related to the target category of the target content.

[0040] When the cover generation entry is triggered, a cover generation strategy related to the target content's target category can be used to intelligently generate a target cover. This can involve generating one or more candidate covers, from which the user selects the target cover.

[0041] Among them, the target cover is generated by the cover generation strategy for the target resources in the target content. The cover generation strategy is derived by analyzing the cover of the content covers of related published content belonging to the target category. The click-through rate of the content covers of related published content is higher than the preset click-through rate threshold.

[0042] For example, when the target content is a target video, the target resource can be a target video frame in the target video, such as a video frame that matches the cover generation strategy. In this case, the image processing method for the target video frame can be determined based on the cover generation strategy (such as how to process the target video frame), and the corresponding image processing prompt words can be constructed based on the image processing method. The target video frame and the image processing prompt words can then be input into the image generation model, and the image generation model can perform image processing on the target video frame based on the image processing prompt words to obtain the processed target video frame. The processed target video frame is the target cover (or candidate cover) for the target video.

[0043] For example, when the target content is target text, the target resource can be the target theme of the target text (such as an object identified from the target text, such as a certain item). In this case, the image generation method for the target theme can be determined based on the cover generation strategy (such as generating an image that matches the target theme, and how to process the image). Based on the image generation method, corresponding image processing prompts can be constructed. The image processing prompts can then be input into the image generation model, and the image generation model can generate the image based on the image processing prompts to obtain the target cover (or candidate cover) for the target text.

[0044] For example, when the target content includes target text and target image, the target resource can be an image in the target image that matches the cover generation strategy (the matching image). In this case, the image processing method for the matching image can be determined based on the cover generation strategy (such as how to process the matching image). Based on the image processing method, the corresponding image processing prompt words can be constructed. The matching image and the image processing prompt words can then be input into the image generation model. The image generation model performs image processing on the matching image based on the image processing prompt words to obtain the processed matching image. The processed matching image is the target cover (or candidate cover).

[0045] The generated candidate covers can be one or more, and different candidate covers are obtained by different cover generation strategies. For example, taking the target content as the target video, a specified number of cover generation strategies can be selected from the cover generation strategies related to the target category (such as cover generation strategies related to the video content of the target video). For each cover generation strategy, the corresponding target video frame is determined, and image processing is performed on the corresponding target video frame according to each cover generation strategy to obtain the candidate covers obtained by image processing of each cover generation strategy. At this time, the candidate covers can be returned to the terminal device for display. Users can view and select a specified candidate cover as the target cover for the content to be published on the content publishing page, or they can also edit the selected target cover on the content publishing page.

[0046] One of the cover generation strategies for a category is obtained by analyzing the cover of related published content in that category (at this time, cover analysis can be performed on only the content cover, or on both the content cover and auxiliary content. The auxiliary content can be the title, content summary, etc. of the related published content, which is not limited here).

[0047] Among them, related published content refers to published content in this category whose click-through rate is higher than the preset click-through rate threshold (such as determining the corresponding category based on related content in the published content, such as video content, text content, etc.). Since the click-through rate of published content is higher than the preset click-through rate threshold, the click-through rate of its cover is also considered to be higher than the preset click-through rate threshold.

[0048] For example, it can acquire video content published under a certain category with a click-through rate higher than a preset click-through rate threshold, and perform cover analysis on the cover of the video content to obtain a cover generation strategy for a category. Thus, a cover strategy library can be built for each category in the pre-built category set to generate target covers for target content.

[0049] It's understandable that, depending on the target content, a single cover strategy library can be used, or separate cover strategy libraries can be employed. For example, for content containing a target video (video content), a cover strategy library could be built specifically for video content within a certain category that has a click-through rate (CTR) exceeding a preset CTR threshold. For content not containing a target video (text / image content), a cover strategy library could be built specifically for text / image content within a certain category that has a CTR exceeding a preset CTR threshold, targeting the target text (or both). Alternatively, regardless of the content type, a shared cover strategy library could be built directly for content within a certain category that has a CTR exceeding a preset CTR threshold. No further limitations are imposed here.

[0050] Taking target classification as an example, at least one cover generation strategy related to the target classification is obtained by clustering the content covers of related published content belonging to the target classification and performing cover analysis on the content covers based on the clustering results. One cover generation strategy corresponds to one clustering result, and any clustering result includes at least one content cover. The content cover refers to the cover with a click-through rate higher than the preset click-through rate threshold, and the related published content refers to the published content belonging to the target classification with a click-through rate higher than the preset click-through rate threshold.

[0051] Optionally, the server can periodically update the cover strategy library, and for a category, it can periodically obtain the latest related published content and perform cover analysis on its content covers to obtain one or more latest cover generation strategies and supplement them in the cover strategy library.

[0052] For example, a cover generation strategy can be derived by an intelligent model through cover analysis of related published content covers. For instance, the cover of related published content (which may also include supplementary content) and cover analysis prompts can be attributed to the intelligent model, which can then summarize one or more cover generation strategies for that category.

[0053] A cover generation strategy can include click motivation (i.e., why users click on the content cover, such as immersive alternatives: fulfilling unfulfilled dreams, e.g., replacing users in experiencing "extreme loneliness"; premium / niche premium: showcasing a unique identity, e.g., showcasing "polar secrets known only to 1% of people"), image processing logic (e.g., structural decisions: single image, collage, subject cutout; post-processing: filters, color correction), and text design logic (copy design: length constraints, tone, popular vocabulary; aesthetics: font type, layout preferences). Subsequently, based on this cover generation strategy, the image processing method for the target cover can be determined (e.g., how to process image structure, how to set displayed text, etc.). In this way, the generated target cover not only contains the target content but also inherits the relevant processing advantages of the cover generation strategy. Since the cover generation strategy is based on analysis of a large number of covers frequently clicked by users, the target cover can match user preferences, thereby attracting users to click and view the published content.

[0054] For example, cover analysis prompts could be as follows: #System Role You are a researcher specializing in creating viral cover strategies, adept at extracting reusable patterns from a large number of content covers.

[0055] #Task Based on the corpus provided below, this is the cover data of the Top 500 most clicked articles in the past 30 days.

[0056] Generate a structured "Cover Trend Report" and output it strictly in JSON format.

[0057] # Input data Category: {category} Data: {top50_cover} / / Format: {rank, ctr, title, cover_image_url, cover_description}, ... #Key Requirements The following three modules are all core outputs. Each module must be listed. Do not merge, omit, or list only the first few items.

[0058] Each strategy / style that identifies a clear pattern in the data will be presented as a separate entry.

[0059] It is better to list more than to omit anything.

[0060] # Output JSON { “category”:{ <category>}” "updated_at": "Generated timestamp" "content_strategy": { / / Solving the "click motivation - why do users click" / / Perform strategy clustering on all content covers, assigning each image to the best-matching strategy, and count the number of hits. "top_motivations": [ { "name": "Policy Name" "description": "A one-sentence description of the essence of this strategy". "hit_count": The number of covers that hit this strategy. "examples": [ { "title": "High Click-Through Rate Title Case 1", "cover_desc": A brief description of the cover, within 20 characters. }, { "title": "High Click-Through Rate Title Case 2" "cover_desc": A brief description of the cover, within 20 characters. }, / / List at least 3 typical cases for each strategy. } ] } / / List all observable strategies until none can be satisfied. } "visual_style": { / / Solve the problem of "Visual Packaging - How to Make It Look Good?", i.e., image processing logic "structure": { "types": ["single image", "puzzle", "combination"], "distribution": Description of the probability of occurrence of each structure type }, "filter_style": { / / List all observable filter / atmosphere styles, each as a separate entry. { "name": "Style name, such as: cool film look". "description": A brief description, within 15 characters. } }, "color_palette": ["Popular Color 1", "Popular Color 2", ...], "decoration_elements": [ / / List all observable decorative features, each as a separate bar { "name": "Decorative name, such as: cool, film-like". "usage": "Common usage, within 15 characters" } ] }, "text_style": { / / Solving the problem of "adding borders to text"—which text to add and how to add them?, i.e., the logic of text design. "has_text_rate": 0.0-1.0, "char_count_range": [min, max], "one": "List all observable copywriting tone styles, such as: conversational / professional / emotional / suspenseful..." "high_weight_words": ["Specific words that frequently appear in the copy and are strongly related to high CTR, list them, the more the better"] "font_style": [ { "name": "Font style name, such as: bold". "example_font": "Specific font name. For example: Founder Lanting Hei / Huawei Bold / Source Han Serif. Leave blank if unsure". "usage_rate": Estimated percentage of usage in the cover copy. } ], "text_position": ["Lists all observable text positioning preferences, such as: top center / bottom left alignment..."] } For example, it can capture the titles and covers of the top 500 video posts with the highest click-through rates in the past 30 days under a certain category (or video posts with click-through rates greater than or above the preset click-through rate threshold), and output a structured report. This structured report includes one or more cover generation strategies. For example, a cover generation strategy includes: (1) "Click Motivation", such as: Cognitive Asymmetry: Challenge common sense (e.g., everyone thinks X place is very expensive, but there is actually a low-cost arbitrage path). Immersive Alternative: Satisfy users' unfulfilled dreams (e.g., replace users to complete the baptism of "extreme loneliness"). Superiority / Niche Premium: Showcase unique identity (e.g., show "the polar secrets known only to 1% of people"). Closed Loop of Gain: Quickly solve anxiety (e.g., solve all the fears of traveling alone in one stop). (2) Image Processing Logic, such as: Structural Decision: Single image, collage (multiple grids), subject cutout. Post-processing: Filters (e.g., high-cold film feel), color correction. Decorative elements: popular elements (hand-drawn arrows, paper wrinkles, film borders, speech bubbles, etc.). (3) Text design logic: for example: copywriting design: length constraints, tone, and popular words. Aesthetics: font type (serif / handwritten / calligraphy / bold, etc.), layout preferences.

[0061] Among them, the clustering results obtained by performing strategy clustering on the content cover can be: (1) Extracting core nouns (such as: yellow and black skin, puffy eyes, 8 am, big brand) from each content cover and / or corresponding title, extracting action instructions (such as: avoid thunder, rush, get it, learn, 3 minutes), extracting emotional words (such as: help, amazing, surprisingly, hottest), extracting visual contrast features (such as: left and right puzzle, arrow pointing, magnifying glass), thereby obtaining the corresponding feature vector: {title keywords: [8 am, change head], action words: [you can do it with your hands], visual elements: [left and right comparison picture], target audience: [none]} (2) Calculating similarity, that is, based on the above features, calculating the "motivational similarity" between all the covers. (3) Algorithm clustering, such as using a clustering algorithm to obtain the clustering results. At this time, a clustering result contains at least one feature vector, and at this time, the corresponding cover generation strategy can be parsed based on at least one content cover corresponding to at least one feature vector in the clustering result (one feature vector corresponds to one content cover). (4) Extract strategy name, click motivation, image processing logic, text processing logic, etc. from the content cover and / or corresponding title to obtain a category-related cover generation strategy.

[0062] S103. In response to a content publishing operation on a content publishing page, publish target publishing content that includes the target content and whose cover is the target cover.

[0063] Selecting a target cover on the content publishing page and then performing the content publishing operation instructs the server to publish target content for the target content, with the target cover serving as the target cover.

[0064] In this embodiment, a content publishing page for publishing target content can be displayed, and a cover generation entry can be shown on the content publishing page. By triggering the cover generation entry, a target cover for the target content can be intelligently generated, facilitating the publication of target content containing the target content and whose cover is the target cover. The generation of the target cover can be achieved by generating an image according to a cover generation strategy related to the target category of the target content. In other words, intelligent image processing can be performed using a cover generation strategy related to the category to which the target content belongs to obtain the target cover. This cover generation strategy is derived from cover analysis of the covers of related published content belonging to the target category. If the click-through rate (CTR) of the content cover exceeds a preset CTR threshold, it's possible to pre-analyze a large number of high-quality content covers (i.e., covers of published content with high CTR) within the target category. This analysis reveals patterns in content covers that attract online users within that category (e.g., image structure, filter style, text layout). A corresponding cover generation strategy can then be developed. Utilizing this strategy for intelligent cover generation yields high-quality covers that appeal to a broad audience and align with the target content. This more intelligent approach to cover selection improves the quality of generated covers, ensuring they are attractive, representative, and meet the needs of the target content and style, thereby enhancing the click-through rate and conversion rate of the published content.

[0065] Based on the above description, this application proposes a cover generation method, which can be executed by the aforementioned electronic device, specifically... Figure 1 The terminal device shown. Please refer to [link / reference]. Figure 4 , Figure 4 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 2 .like Figure 4 As shown, the cover generation method of this application embodiment may include the following steps: S201. Display the content publishing page used to publish the target video.

[0066] The target content can be a target video, in which case the uploaded target video will be displayed on the content publishing page. Additionally, the content publishing page includes an entry point for generating a cover image.

[0067] The content publishing page allows users to view the current content cover (e.g., the first frame of the target video by default). It can also be used to preview the target content to be published, allowing users to check the publishing progress.

[0068] For example, the content publishing page displays detail page preview controls and cover preview controls. When the detail page preview control is selected, the content publishing page displays a preview of the content detail page, and the target video is displayed on the content detail page (i.e., previewing the target published content in the form of a content detail page). When the cover preview control is selected, the content publishing page displays a preview of the content recommendation page; the content recommendation page is used to display content cards, and the content cards display the cover determined by the content publishing page (i.e., previewing the content cover in the form of a content recommendation page). It can be understood that when the content cover changes, the cover displayed on the previewed content recommendation page will also change accordingly.

[0069] S202. In response to a trigger operation on the cover generation entry, display the target cover generated according to the cover generation strategy related to the target video's target category.

[0070] The content publishing page is used to publish target content that includes the target video and whose cover image is the target video cover image.

[0071] The target cover is obtained by image processing of the target video frame in the target video through a cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of related published content is higher than the preset click-through rate threshold.

[0072] For example, in response to a trigger operation on the cover generation entry, the target cover generated according to the cover generation strategy related to the target category of the target video can be displayed. Alternatively, in response to a trigger operation on the cover generation entry, at least one candidate cover generated for the target video can be displayed. A candidate cover is generated according to a cover generation strategy, and the cover generation strategy corresponding to each candidate cover is different. In response to a selection operation on at least one candidate cover, the selected candidate cover is determined as the target cover.

[0073] For example, the target category of the video can be determined, and cover generation strategies related to the target category can be obtained from the cover strategy library. A specified number of cover generation strategies can then be selected, and image processing can be performed according to each strategy to obtain a specified number of candidate covers. Users can then choose the target cover from these candidate covers as the content cover.

[0074] Taking a cover generation strategy as an example, the strategy involves processing target video frames that match the strategy to obtain candidate covers. Specifically, it identifies one or more target video frames that semantically match the strategy, and processes these frames to generate a candidate cover. (For a single target video frame, image processing is performed directly on that frame; for multiple target video frames, they can be stitched together into a single image before processing, or processed and then stitched together into a single image.) Each cover generation strategy is different, and the matched target video frames can also be different.

[0075] The cover generation strategy used for image processing can be randomly obtained from cover generation strategies related to the target classification, or it can be obtained based on the video content of the target video. For example, the video summary information of the target video can be obtained, and a specified number of cover generation strategies can be obtained based on the video summary information. Subsequently, the target video frames can be obtained based on the cover generation strategies for image processing.

[0076] Optionally, the target video frame can be matched based on the cover generation strategy, or it can be specified by the user. For example, the user can specify one or more video frames as the target video frame and perform image processing on the specified target video frame through the cover generation strategy.

[0077] For example, in response to a video frame selection operation for a target video, the selected video frame in the target video is determined as the selected video frame; in response to a trigger operation for a cover generation entry, after performing image processing on the selected video frame according to the cover generation strategy related to the target classification of the target video to obtain the target cover (i.e., one or more candidate covers), the target cover is displayed.

[0078] The cover generation strategy used to process the selected video frames can be obtained in the manner described above (such as random acquisition or acquisition based on the video content of the target video), or it can be based on the selected video frames. For example, it can be to acquire a specified number of cover generation strategies that semantically match the selected video frames for image processing. No limitation is made here.

[0079] For example, in response to a trigger operation targeting the cover generation entry, displaying the target cover generated according to the cover generation strategy related to the target video's target category could involve: generating a cover generation request for the target video in response to the trigger operation targeting the cover generation entry; sending the cover generation request to the server so that the server can perform video analysis on the target video to obtain the target video's target category, video summary information, and frame attribute information for each video frame in the target video; determining the cover generation strategy associated with the video summary information from at least one cover generation strategy related to the target category; selecting target video frames from the video frames of the target video whose frame attribute information matches the associated cover generation strategy; when determining the image processing strategy for the target video frame based on the associated cover generation strategy; inputting the target video frame and the image processing prompts constructed for the image processing strategy into the image generation model; and having the image generation model perform image processing on the target video frame based on the image processing prompts to obtain the target cover; and displaying the target cover on the content publishing page upon receiving the target cover returned by the server.

[0080] Therefore, the server pre-builds a cover strategy library, and for each category in the cover strategy library, it obtains related published content with a click-through rate higher than a preset click-through rate threshold, performs cover analysis on the content covers of the related published content for each category, and summarizes at least one cover generation strategy related to each category.

[0081] Furthermore, after receiving a cover generation request and performing video analysis, the server retrieves at least one cover generation strategy related to the target classification from the cover strategy library, and determines at least one cover generation strategy associated with the video summary information from the at least one cover generation strategy related to the target classification. It then selects target video frames from the video frames whose frame attribute information matches each associated cover generation strategy. When determining each image processing strategy for the target video frame based on each associated cover generation strategy, the server inputs the target video frame and each image processing prompt word constructed for each image processing strategy into the image generation model. The image generation model then performs image processing on the target video frame based on each image processing prompt word to obtain at least one candidate cover.

[0082] The server needs to perform video analysis on the target video content to determine the target video's category, video summary information, and frame attribute information. Based on the target category, at least one relevant cover generation strategy can be obtained, and the matching cover generation strategy can be obtained using the video summary information. For each cover generation strategy and each video frame's frame attribute information, the target video frame corresponding to each cover generation strategy is determined (a cover generation strategy and its corresponding target video frame's frame attribute information have semantic matching). The cover generation strategy can then be used to determine the image processing method for the target video frame. After constructing corresponding image processing prompts, the image generation model performs image processing on the target video frame based on the image processing prompts, obtaining candidate covers that can serve as the target cover.

[0083] Among these, video analysis of the target video, selection of cover generation strategy, and construction of image processing prompts can be achieved by intelligent models based on relevant prompts.

[0084] Among them, video analysis of the target video can be performed by scanning the entire target video and extracting frames to identify visual features (subject, composition, lighting) and semantic information (ASR script, keywords) to explore the potential of every part of the video material. Output: (1) Overall video understanding: the target category, video summary information, i.e., core summary (core_topic). Among them, the prompt words used for video understanding can be: #System Role You are a professional video content analyst, skilled at understanding video content from both visual and semantic dimensions.

[0085] #Task Based on the following input, provide an overall understanding of this video and output it strictly in JSON format without any other content.

[0086] #enter ASR script: {asr_script} Author's pre-filled title (can be empty): {author_title} # Output JSON format { "category": "category" "core_topic": "Core summary, within 15 words, clearly explaining what this video is about." "emotion_tone": "Overall emotional tone, choose one from the following: healing / inspirational / curious / funny / suspenseful / professional information / emotional resonance / longing / indignation". "key_hook": "What is the most crucial hook element in this video, within 20 characters?" "asr_keywords": ["High-weight keywords extracted from the script, up to 8"] } (2) Frame attribute construction: Each frame is labeled with an attribute tag, i.e., the video frame attribute information of each video frame is obtained. For example, Frame types: emotional peak | product close-up | character appearance | action climax | scene aesthetics... Emotional tone: Warm | Tense | Humorous | Professional | Inspirational | Curious | Suspenseful | Attractive | Indignant | Longing... Is there anyone: Expert | Ordinary Person | Celebrity / KOL | Baby | No Person... Product availability: Product Name / No Product... Does it have subtitles? What are the keywords for the subtitles? / No subtitles... What is the most appealing element of the image that makes you want to click? (In 15 words or less)

[0087] The prompt words used for constructing frame attributes can be: #System Role You are a professional cover frame analyst, skilled at extracting potential cover signals from single frames.

[0088] #Context / / (i.e., the information summarized in the video above) Video elements: {category} Core Summary: {core_topic} Mood tone: {emotion_tone} #Task Extract the structured attributes from the frame below and output it strictly in JSON format without any other content.

[0089] #Analysis Dimensions Explanation frame_type: What is the most outstanding visual value of this frame? Summarize it freely in 2-4 words, without being limited to fixed categories.

[0090] (Reference factors: emotional value, product symbols, character appearance, action highlights, scene aesthetics, daily life, key points of subtitles... but not limited to these, make your own judgment based on the actual content of the image) Emotion: What is the emotional tone conveyed by this frame? Summarize it freely in 2-4 words.

[0091] (Reference themes: heartwarming, tense, humorous, longing, inspirational, curious, suspenseful, indignant... but not limited to these. Please make your own judgment based on the actual emotions conveyed in the image.) person_type: The type of person in the picture. Summarize it freely in 3-6 words.

[0092] (Reference areas: ordinary women, expert commentators, celebrities / KOLs, cute babies, no people... but not limited to these) subject_position (subject position, used for subsequent copywriting logic, choose one from the following): Center | Left | Right | Top | Bottom | Full Screen # Output JSON { "frame_id": {frame_id}, "timestamp_sec": {timestamp_sec}, "frame_type": "A brief summary, 2-4 words". "emotion": "Summarize freely, 2-4 words". "person_type": "Free summary, 3-6 words, or "no person"". "product": "Product name or "No product" "subject_keywords": "Subtitle keywords or 'subtitle'", "subject_position": "Center | Left | Right | Top | Bottom | Full Screen" "dominant_colors": ["dominant color 1", "dominant color 2"], "click_magnet": "The element in this frame that most attracts the user's desire to click, describe it freely in 15 characters." "cover_potential_score": 0-10, "cover_potential_reason": "Reason for rating, within 20 characters" } In this process, after determining the target category of the target video, the image processing method for the target video frame can be determined based on the cover generation strategy related to the target category.

[0093] For example, the video analysis results for the target video can be multimodally aligned with the cover strategy library to obtain: (1) the matching cover generation strategy, i.e. which cover generation strategies the target video can match. For example, video A can match the cover generation strategies of "One-stop strategy to solve anxiety" and "extreme loneliness". (2) find the target video frame, i.e., first perform semantic alignment, i.e., take the cover generation strategy (such as "extreme loneliness") and find the "base map" that can carry the logic in the full frame material, i.e. the video frame whose frame attribute information has semantic matching with the cover generation strategy. Then perform video frame screening, i.e. judge the potential, i.e. see if the matched video frame can be designed in the later stage to meet the aesthetic trend, i.e. the intelligent model evaluates whether the screened video frame can be used as the target video frame. At this time, one or more target video frames can be determined for the cover generation strategy. (3) image processing, i.e. determine the processing methods of filter, color adjustment, collage / cutout, decoration, etc. (4) text processing, i.e. determine the processing methods of text content, font style, and position description (requiring avoidance of the main body). At this point, the image processing and text processing are determined by the cover generation strategy for the target video frame.

[0094] For example, the prompt words to determine the image processing method could be: #System Role You are a cover image strategy overview, skilled at combining content insights, visual aesthetics, and copywriting to select the most potentially viral cover designs for various types of videos. You deeply understand that the platform, as the core of a "talent selection and product recommendation" community, pursues authentic and engaging user connections regardless of the type of notes.

[0095] # Enter the above text ## Overall video understanding / / i.e., the information summarized in the video above - Video Category: [[category]] - Core Summary: [[core_topic]] - Mood tone: [[emotion_tone]] - Core hooks: [[key_hooks]] - High-weight keywords: [[key_words]] - Initial resolution / aspect ratio: [[frame_dimensions]] / / Format example: "1920x1080 (16:9)" or "1080x1920 (9:16)" ## Full Frame Attribute List / / i.e., the frame attribute information mentioned above [[frame_attributes_list]] / / Format: [[frame_id, timestamp_sec, frame_type, emotion, person_type, / / subject_position, click_magnet, cover_potential_score, ...]] ## Cover Strategy Library - Content strategy list: [[context_strategy_top_motivations]] - Mainstream visual structure: [[visual_structure]] - Popular filter styles: [[visual_style_filter_style]] - Popular color palette: [[visual_style_color_palette]] - Popular decorative elements: [[visual_style_decoration_elements]] - Text usage rate: [[visual_style_has_text_rate]] - Copywriting word count range: [[visual_style_char_count_range]] - Text style and tone: [[visual_style_tone]] - High-frequency words: [[visual_style_high_weight_words]] - Mainstream font styles: [[visual_style_font_style]] - Font position preference: [[visual_style_text_position]] # Task Complete the following four steps to generate three differentiated cover designs, and output them strictly in JSON format.

[0096] ##

Step 1: Content Strategy Matching

[0097] - Requirements: The click motivations of the three strategies must not overlap and must cover different user psychology.

[0098] ##

Step Two: Semantic Alignment Filtering

[0099] - rule: - A solution can use a single frame (single image cover) or multiple frames (puzzle / composite cover); - The three schemes must select different frames, and the frame_id cannot be repeated; - Visual inspection: Does the color tone match the travel photo? Is the portrait clear (e.g., face)? Does the position of the main subject support the placement of the text? Is the overall coordination of multiple frames combined?

[0100] ##

Step 3: Image Processing Solution Design

[0101] - Canvas size adjustment: [[frame_dimensions]] determines whether the original canvas is horizontal (e.g., 16:9) or the final generated cover ratio is 4:3; if the original canvas is vertical (e.g., 9:16), the final generated cover ratio is 3:4.

[0102] - Structure: Determine the single image / puzzle / combined image. For a single image, explain how to crop from the original frame to the target aspect ratio while ensuring the main subject is not cut off; for a puzzle, explain the arrangement and segmentation of multiple frames on the target aspect ratio canvas.

[0103] - Filters: Based on the range of popular filters in the trend library, specific filter options are provided.

[0104] - Color adjustment: Combining the trend library's popular color palette, it provides specific directions for color tone processing (such as increasing or decreasing color temperature, saturation, and transparency).

[0105] - Decorative elements: Select elements from the trend library that match the style and theme of this strategy, and explain their placement and style.

[0106] - Special red lines for merchants: If this video involves commercial brands / stores, it must strictly avoid any traces of e-commerce marketing. Absolutely no "promotional prices, discounts, price reductions, or official brand logos" may be added. Avoid any human-like promotional elements or bordering on e-commerce promotional fads. All color schemes and decorations must serve a "lifestyle" and "everyday sharing atmosphere."

[0107] ##

Step 4: Cover Copy Design

[0108] - If text is required (has_text = true): - Copywriting: Based on the content strategy of this solution, combined with high-weight keywords in the trend library, semantic tone and style, and word limit, create the title; - Font: Select the font style that best matches the visual style of this scheme from the Trend Library font style list; - Layout position: Select from the biased position preferences in the trend library, and actively avoid the area where the main body is located. Explain the specific placement position; - Special Red Line for Merchants: If this video involves commercial brands / stores, the copy must not contain brand names, specific prices, discount figures (such as 9.99, 30% off), or e-commerce sales pitches such as "lowest price online," "buy one get one free," or "limited-time special offer." The copy must be conversational and persuasive, sharing genuine experiences, and highlighting pain points / scenarios.

[0109] # Output JSON json { "plan_id": "A" "strategy_name": "The name of the content strategy to be selected". "strategy_reason": "Why this strategy is suitable for this video, in two words or less". "frame_id": "The ID of the selected frame, ["the ID of the selected frame"]", "timestamp_sec": "[seconds, seconds2]", "frame_select_reason": "Reason for selecting this frame, within 25 characters". "image_processing": { "canvas_ratio": "4:3 or 3:4 (strictly determined by the original frame aspect ratio)" "structure": "Single image / jigsaw puzzle / combined image". "layout_note": "When using a single image, explain how to crop it to the target aspect ratio without cutting into the main subject; when using a mosaic, explain the arrangement and storyboarding of multiple frames on the target aspect ratio canvas." "filter_style": "The name of the selected filter style", "color_grading": "Specific color grading direction, such as color temperature and chroma adjustment, within 5 characters". "decoration": "Decoration item name, location, and style description" }, "has_text": true / false "copy": { "title": "Main title, null if none". "subtitle": "Subtitle, null if none". "font_style": "The name of the selected font style". "position": "Establish a position within the main text; place the text within [specific margins]". } } / / Note: If has_text is false, then the "copy" field is set to null, and "no_text_reason" is added: a reason for not adding text, within 15 characters. Once the image processing method is determined (such as the image processing scheme design and cover text design mentioned above), image processing prompts are constructed that can be input into the image generation model, i.e., outputting multiple sets of raw image instructions. For example, output: raw image instruction; underlying image: target video frame ID; image processing: filter, color correction, collage / cutout, decorations; text processing: text content + font style + location description (requiring avoidance of the main subject).

[0110] For example, the prompts used to construct image processing prompts could be: #System Role You are the Executive Director of Cover Visuals, responsible for translating decision-making solutions into actionable instructions that can be directly delivered to designers or image generation tools.

[0111] # Inject current scheme context / / Image processing method from the above output Plan ID: {plan.plan_id} Selected frames: {plan.frame_ids} (timestamps: {plan.timestamp_secs}) Image processing solutions: Structure: {plan.image_processing.structure} Arrangement logic: {plan.image_processing.layout_note} Filter style: {plan.image_processing.filter_style} Color grading direction: {plan.image_processing.color_grading} Decoration: {plan.image_processing.decoration} Copywriting processing solution: Does it contain text: {plan.has_text} Title of the document: {plan.copy.title} Subtitle of the text: {plan.copy.subtitle} Font style: {plan.copy.font_style} Font position: {plan.copy.position} #Task The above scheme is converted into structured raw image execution instructions, which are output strictly in JSON format.

[0112] Output JSON { "plan_id": "A" "image_prompt": { "base_frame_ids": ["baseframe IDs"], "image_processing": { "structure": "Single image | collage | image cutout compositing", "layout_note": "Describes the frame arrangement logic when creating a jigsaw puzzle; null when creating a single image". "filter": "Filter style name and detailed description, within 30 characters" "color_grading": "Specific color grading direction, such as: lowering the color temperature to 4500K, increasing the cyan channel", "decoration": ["Decoration name: location and style description"] }, "text_processing": { "title": "Main title text" "subtitle": "Subtitle text; null indicates omission". "font_style": "font style name", "position": "The main content is in [position], and the copy is placed in [specific, less prominent position]." } / / When has_text=false, text_processing is null. } } It is understood that the above prompts are merely examples and are not limited here; any prompts that can achieve the relevant logic are acceptable.

[0113] For example, such as Figures 5a-5e As shown, Figures 5a-5e This application provides a schematic diagram of a cover selection scenario; wherein, in Figure 5a On the content publishing page 50a, the target video to be uploaded can be displayed, and the content cover (the first frame is displayed by default) can also be viewed. Additionally, examples of high-quality cover images can be viewed to help users understand the cover images preferred by online users. Titles and body text can be entered on the content publishing page. Optionally, when a target video is uploaded, the title and body text can be intelligently generated based on the target video.

[0114] The content publishing page displays a details page preview control 50b and a cover preview control 50c.

[0115] When the preview control on the details page is selected, the content details page 50d of the target content to be published can be previewed on the content publishing page. At this time, the target video (such as displaying a video frame in the target video) can be displayed on the content details page to view the publishing effect.

[0116] When the cover preview control is selected, the content recommendation page can be previewed on the content publishing page. The content recommendation page can display content cards 50e, such as the content card of the target content to be published. The content card is used to display the content cover determined by the content publishing page to view the display effect of the cover.

[0117] Among them, Figure 5b In the content publishing page 51a, a cover generation entry 51b is displayed. Triggering this entry instructs the server to generate at least one candidate cover 51c for the target video. During candidate cover generation, the content publishing page can display progress indicator text 51d to allow users to perceive the generation progress. For example, the first stage is video understanding, which involves analyzing and obtaining video summary information and frame attribute information. The second stage is cover strategy library matching, which involves matching cover generation strategies based on video summary information and matching target video frames based on frame attribute information and cover generation strategies. The third stage is cover scheme design, which involves determining image processing methods and constructing image processing prompts to generate candidate covers, etc.

[0118] It is understandable that at different stages, displaying production progress prompts that match the production progress is necessary so that users can stay informed.

[0119] For example, in the second stage, the displayed progress notification text could be: Status phrases (played in sequence): - Cover strategy library search in progress... - The {category} category has been matched for you... The completed text will be displayed (assembled based on content_strategy.top_motivations): There are {N} types of popular cover strategies in this category recently, which are: [{top_motivations[0].name}]— {top_motivations[0].description}; [{top_motivations[1].name}]— {top_motivations[1].description}; [{top_motivations[2].name}]— {top_motivations[2].description}.

[0120] After obtaining the cover generation strategy to be used, the displayed generation progress prompt text can be: Status phrases (carousel) - Performing multimodal alignment between [Video Content Understanding] and [Cover Strategy Library]... - Cover design in progress... Once the cover design is complete, the displayed progress notification text can be: Upon completion, the following text will be displayed (based on the plans array): Three differentiated cover designs have been generated for you: Plan A[{plan_A.strategy_name}] — {plan_A.strategy_reason} Plan B[{plan_B.strategy_name}] — {plan_B.strategy_reason} Plan C[{plan_C.strategy_name}] — {plan_C.strategy_reason} In this process, after the server generates at least one candidate cover using an image generation model, it returns the result to the terminal device. The terminal device can then display at least one candidate cover on the content publishing page, and the user can also choose to regenerate it.

[0121] Among them, Figure 5c In the content publishing page 52a, at least one candidate cover image is displayed, and the user can select one as the target cover. For example, when a candidate cover image 52b is selected, a confirmation control 52c can be displayed. After triggering the confirmation control again, the user can confirm that the candidate cover image is the target cover. At this time, the content cover image 52d displayed on the content publishing page can be updated to the target cover image. In addition, the cover image displayed in the content card on the preview content recommendation page is also updated to the target cover image (candidate cover image 52b). At this time, if a content publishing operation is performed through the content publishing page, the corresponding target published content includes the target video, and the content cover image is the target cover image.

[0122] Among them, Figure 5d In the process of generating candidate covers, it can be: (1) Obtain the associated published content under the target category, cluster the content covers of the associated published content, and obtain the clustering results (such as clustering results 1-4). Each clustering result includes at least one content cover. At this time, summarize the content covers in a clustering result and obtain the corresponding cover generation strategy (such as cover generation strategy 1-4). Based on the cover generation strategy corresponding to each category, construct a cover strategy library.

[0123] (2) When the server receives a cover generation request for a target video, it can perform video analysis on the target video to obtain the target classification, video summary information, and frame attribute information of each video frame in the target video. Among them, based on the target classification of the target video, the server can obtain a cover generation strategy associated with the video summary information from at least one cover generation strategy related to the target classification in the cover strategy library (such as cover generation strategies 1-3). At this time, the server can select target video frames whose frame attribute information semantically matches the associated cover generation strategy for each cover generation strategy (such as video frame a matched by cover generation strategy 1, video frame b matched by cover generation strategy 2, and video frame c matched by cover generation strategy 3).

[0124] In this process, the corresponding image processing strategy (i.e., image processing method) can be determined based on each cover generation strategy. For example, image processing strategy 1 for video frame a can be determined based on cover generation strategy 1, image processing strategy 2 for video frame b can be determined based on cover generation strategy 2, and image processing strategy 3 for video frame c can be determined based on cover generation strategy 3.

[0125] Image processing prompts can be constructed based on image processing strategies. For example, image processing prompt 1 can be determined based on image processing strategy 1, image processing prompt 2 can be determined based on image processing strategy 2, and image processing prompt 3 can be determined based on image processing strategy 3.

[0126] Specifically, image processing prompt 1 and video frame a can be input into the image generation model to obtain candidate cover 1; image processing prompt 2 and video frame b can be input into the image generation model to obtain candidate cover 2; image processing prompt 3 and video frame c can be input into the image generation model to obtain candidate cover 3.

[0127] Optionally, in Figure 5e In this context, the target video frame can be intelligently selected by the smart model based on the cover generation strategy, or it can be specified by the user. For example, the content publishing page 53a displays a video frame selection entry 53b. By triggering the video frame selection entry, a video frame selection page 53c is displayed, which shows video frames from the target video. At this time, the user can specify one or more video frames as the selected video frames (i.e., target video frames). The selected video frames 53d can then be viewed on the content publishing page. Furthermore, the video frame selection entry can be changed to a reselection entry 53e to reselect video frames. When the cover generation entry is triggered, the server can be instructed to perform image processing on the selected video frame 53d to obtain at least one candidate cover obtained from the selected video frame 53d.

[0128] In this embodiment, a content publishing page for publishing the target video can be displayed, and a cover generation entry can be shown on the content publishing page. By triggering the cover generation entry, a target cover for the target video can be intelligently generated, facilitating the publication of target content containing the target video and whose cover is the target cover. The generation of the target cover can be achieved by processing the target video frames in the target video according to a cover generation strategy related to the target video's target category. A target video frame refers to a video frame that matches the cover generation strategy. In other words, the cover generation strategy related to the target video's category can be used to perform intelligent image processing on video frames that semantically match the cover generation strategy, and the processed target video frame can be used as the intelligently generated target cover. This cover generation strategy involves processing the cover of the content published in the relevant category. The analysis of cover images reveals that the click-through rate (CTR) of related published content covers exceeds a preset CTR threshold. This means that we can analyze a large number of high-quality content covers (i.e., content covers of published content with high CTR) within the target category in advance to obtain patterns in content covers that attract online users within that target category (such as image structure, filter style, text layout, etc.). This leads to a corresponding cover generation strategy. By using the cover generation strategy for intelligent cover generation, we can obtain high-quality covers that conform to popular preferences and fit the video content. This allows for a more intelligent cover determination method to improve the quality of cover generation, making the determined covers attractive and representative, meeting the needs of the video content and style, and thus improving the video's click-through conversion rate (i.e., the probability of users clicking on the content cover to view the published content).

[0129] Based on the above description, this application proposes a cover generation method, which can be executed by the aforementioned electronic device, specifically... Figure 1 The terminal device shown. Please refer to [link / reference]. Figure 6 , Figure 6 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 3 .like Figure 6 As shown, the cover generation method of this application embodiment may include the following steps: S301. Display the content publishing page used to publish the target video.

[0130] S302. In response to a trigger operation on the cover generation entry, display at least one candidate cover generated according to a cover generation strategy related to the target video's target category. For specific implementation details of steps S301-S302, please refer to the relevant descriptions in the above embodiments.

[0131] S303. In response to a triggered operation on a cover selection entry on a content publishing page, a cover selection area is displayed on the content publishing page, and the target cover selected from at least one candidate cover and / or the uploaded cover uploaded by the target account are displayed in the cover selection area.

[0132] Users can edit the cover of the current content on the content publishing page. For example, the content cover may be a target cover selected from at least one candidate cover. In response to an edit trigger operation on the target cover, a cover editing page for the target cover can be displayed; in response to a prompt word entry operation on the cover editing page, the edited target cover can be displayed. The edited target cover, determined by the cover editing page, is displayed on the content publishing page. The edited target cover is obtained by image processing of the target cover using the editing prompt word corresponding to the prompt word entry operation.

[0133] For example, when a content cover (such as a target cover) is triggered, a cover editing entry can be displayed. This entry will then open the cover editing page, where users can edit the cover. This could involve using prompts to instruct an image generation model to process the target cover image, or adding templates, filters, or text to the target cover directly on the cover editing page. No further limitations are specified here. The edited content cover will be updated on the content publishing page accordingly.

[0134] Optionally, for the target content to be published, the content cover is usually a set image. To enrich the content display, multiple images can be selected as content covers. In this case, after the target content is published, when pushing the target content to the terminal device, one image can be randomly selected as the content cover currently displayed on the terminal device. For example, if the publishing account selects images 1-3 as content covers for publication, for user A, when displaying the content recommendation page, if the recommended publications displayed on the content recommendation page are the target content, then the content recommendation page will display a content card for the target content. The content cover displayed on this content card can be an image selected from images 1-3.

[0135] For example, an image could be randomly selected from images 1-3 as the cover image currently displayed on the content card, or an image could be adaptively selected based on the current user's (i.e., the current push account) account information. For instance, an image could be intelligently selected from images 1-3 based on the cover images the user recently clicked on, and then used as the cover image currently displayed on the content card.

[0136] Optionally, the server can also notify the click-through rate of each image and use the image with the highest click-through rate as the cover image for the target content. Subsequently, when displaying the target content, this image with the highest click-through rate will always be used as the cover image. This allows users to set multiple covers when publishing and select any one to display when pushing to other accounts. The cover image with the highest click-through rate over a period of time (i.e., the one with the best display effect and most users' preference) will be used as the final content cover image. This improves the flexibility of cover display and can achieve better interactive conversion.

[0137] For example, the content publishing page includes a cover selection entry. When the cover selection entry is selected, a cover selection area is displayed on the content publishing page. This area is used to display one or more target covers and / or one or more uploaded covers from the target account. The target cover is selected from at least one candidate cover generated for the target video. Each candidate cover is generated according to a cover generation strategy, and each candidate cover corresponds to a different cover generation strategy.

[0138] In other words, users can not only select one or more cover images from the candidate covers to add to the cover selection area, but also upload their own images to add to the cover selection area. Optionally, users can also select specific video frames from the video frames to add to the cover selection area.

[0139] Optionally, the covers displayed in the multi-select area can also be edited. For example, triggering the selection of covers in the multi-select area can display the cover editing entry, which will then display the cover editing page. When the edited cover is obtained, it can be updated in the multi-select area.

[0140] S304. In response to the content publishing operation on the content publishing page, determine the target cover and / or uploaded cover displayed in the cover multi-select area as the reference cover, and publish the target published content that includes the target video and whose content cover is the reference cover.

[0141] It is understandable that the cover displayed in the cover multi-select area can be used as a reference cover. After the content publishing operation is performed, the server publishes the target content containing the target video and whose content cover is the reference cover. When pushing the target content, any cover is obtained from the reference covers and displayed as the current content cover.

[0142] For example, such as Figures 7a-7c As shown, Figures 7a-7c This application provides a schematic diagram of a cover editing scenario; wherein, in Figure 7a In the content publishing page 70a, a content cover 70b can be displayed. This cover can be selected from at least one candidate cover. Triggering the content cover will display a cover editing entry 70c, which in turn will display a cover editing page 70d. On this page, users can perform cover editing operations (such as entering editing prompts, which the service then inputs into an image generation model. The model processes the content cover image using the prompts to produce the edited content cover 70e). The cover editing page allows users to view both the original and edited content covers. Users can choose whether to use the edited cover; for example, after confirming the edited cover, the content publishing page can be updated to display the edited cover.

[0143] Among them, Figure 7b In the content publishing page 71a, a cover selection entry 71b can be displayed. When the cover selection entry is selected, a cover selection area 71c can be displayed on the content publishing page (this can be displayed either when the cover selection entry is selected or when any cover is selected; there is no limitation here. For example, the cover selection area can be displayed when uploading any image or selecting any candidate cover). At this time, an image can be selected to upload on the content publishing page and displayed in the cover selection area as the uploaded cover (e.g., displaying the cover upload entry 71d), or a specific candidate cover can be selected from at least one candidate cover and displayed in the cover selection area as the target cover, or a specific video frame can be selected from the video frames of the target video and displayed in the cover selection area (e.g., displaying the video frame selection entry 71e). In this way, after the content is published, the server can be instructed to publish the target content based on the target video, text (such as title, body text, etc.), and the cover in the cover selection area.

[0144] For example, selecting a target cover from candidate covers and adding it to the cover multi-select area could be done by adding the triggered candidate cover to the cover multi-select area when any candidate cover is triggered. Alternatively, when any candidate cover 72a is triggered, a cover confirmation entry 72b can be displayed, and when the cover confirmation entry is triggered again, the triggered candidate cover can be added to the cover multi-select area.

[0145] Among them, Figure 7c In the multi-select area, covers can also be edited. For example, when any cover 72a in the multi-select area is selected, the cover editing entry 72c can be displayed, which will take you to the cover editing page for that cover.

[0146] In this embodiment, a content publishing page for publishing the target video can be displayed, and a cover generation entry can be shown on the content publishing page. By triggering the cover generation entry, a target cover for the target video can be intelligently generated, facilitating the publication of target content containing the target video and whose cover is the target cover. The generation of the target cover can be achieved by processing the target video frames in the target video according to a cover generation strategy related to the target video's target category. Since the cover generation strategy is derived from cover analysis of content covers belonging to the target category, it is based on the regularity of content covers that attract online users within that target category (e.g., image structure, filter style, text). (Layout, etc.) By using cover generation strategies for intelligent cover generation, high-quality covers that conform to popular preferences and fit the video content can be obtained. In addition, for users, there is no limit to the number of covers. For example, multiple covers can be selected as content covers for publication. When recommending the target published content to other users later, one cover can be selected from multiple covers for display, so that they can compete for display in the homepage recommendation scenario. This not only improves the quality of cover generation through a more intelligent cover selection method, making the selected covers attractive and representative, thereby improving the video's click-through rate, but also increases the diversity of cover display through multiple covers, thereby determining the cover that best meets the online user's preferences and further improving the click-through rate.

[0147] Based on the above description, this application proposes a cover generation method, which can be executed by the aforementioned electronic device, specifically... Figure 1 The terminal device shown. Please refer to [link / reference]. Figure 8 , Figure 8 A flowchart illustrating a cover generation method provided in this application embodiment. Figure 4 .like Figure 8 As shown, the cover generation method of this application embodiment may include the following steps: S401. Display the content publishing page used to publish the target text.

[0148] The content publishing page includes a cover generation entry. When the content publishing page contains only target text, this cover generation entry can be used to generate at least one candidate cover for the target text. The target cover can be selected from at least one candidate cover. Thus, when publishing target content through the content publishing page, the target content can include the target text, and the content cover will be the selected target cover.

[0149] S402. In response to a trigger operation on the cover generation entry, display the target cover generated according to the cover generation strategy related to the target text's target category.

[0150] The target cover is obtained by generating an image of the target theme of the target text using the cover generation strategy.

[0151] The server can determine the cover generation strategy related to the target category based on the category to which the target text belongs, select a specified number of cover generation strategies associated with the target text, and determine the image processing method for the target theme based on the cover generation strategy. Then, it can determine the image generation prompt words based on the image processing method and the target theme. The image generation prompt words are used to indicate the generation of an image whose content conforms to the target theme, and the generated image is processed according to the image processing method to obtain candidate covers.

[0152] The target theme is the main information extracted from the target text, such as a person, an object, or a holiday. The image generation model can be instructed to first generate a suitable image based on the target theme, and then process this generated image according to the image processing method indicated by the cover generation strategy. One cover generation strategy can correspond to one candidate cover.

[0153] The cover generation strategy is derived by analyzing the cover images of related published content belonging to the target category. The click-through rate of the cover images of related published content is higher than the preset click-through rate threshold.

[0154] When the displayed content publishing page is used to publish target text and at least one target image, a target cover (such as generating at least one candidate cover) can be generated for the target text and at least one target image. The content publishing page is used to publish content containing the target text, with the content cover being the target cover. The at least one target image refers to an image uploaded by a user.

[0155] The target cover is obtained by image processing of the matching image in the at least one target image through the cover generation strategy, and the matching image is the image that matches the cover generation strategy.

[0156] In other words, a cover generation strategy related to the target text can be determined based on the target text's classification, and a specified number of cover generation strategies associated with the target text can be selected. Then, images that semantically match each cover generation strategy can be identified from at least one target image (by determining the image attribute information of the target image and matching it with the cover generation strategy). The matched images are then processed according to the cover generation strategy to obtain candidate covers.

[0157] For example, such as Figure 9 As shown, Figure 9 This is a schematic diagram illustrating a smart cover generation scenario provided in an embodiment of this application. The content publishing page 90a displays an uploaded target image 90b and entered target text 90c. When the cover generation entry 90d on the content publishing page is triggered, it can intelligently generate at least one candidate cover based on the target text and target image, which the user can then select. For example, after obtaining the cover generation strategy through the target text, image processing can be performed on the matched target image to obtain candidate covers.

[0158] Optionally, candidate covers can be generated by specifying images from the target images. For example, one or more images can be selected from the target images displayed on the content publishing page by triggering the image selection entry to generate a smart cover.

[0159] Optionally, the content publishing page includes a cover selection entry 90e. After triggering this entry, multiple images can be selected as reference covers to publish target content containing the target text and at least one target image, with the content cover being one of the reference covers. This can involve selecting images from candidate covers, uploading images manually, or selecting images from target images. No limitation is imposed here.

[0160] In this embodiment, a content publishing page for publishing target text can be displayed, and a cover generation entry can be shown on the content publishing page. By triggering the cover generation entry, a target cover for the target text can be intelligently generated, facilitating the publication of target content containing the target text and whose cover is the target cover. The generation of the target cover can be achieved by image generation according to a cover generation strategy related to the target text's category. In other words, intelligent image processing can be performed using a cover generation strategy related to the category to which the target content belongs to, resulting in the target cover (e.g., intelligently generating a suitable cover that conforms to popular preferences based on user-inputted text and title). This cover generation strategy is achieved by... The analysis of cover images in content published within the target category reveals that, based on a large number of high-quality content covers (i.e., covers of high-click-through-rate published content) in that target category, we can pre-analyze the patterns in cover images that attract online users within that target category (such as image structure, filter style, and text layout). By using cover generation strategies for intelligent cover generation, we can obtain high-quality covers that conform to popular preferences and fit the target content. This more intelligent cover determination method can improve the quality of cover generation, making the determined covers attractive and representative, meeting the needs of the target content and style, thereby improving the click-through rate and conversion rate of the target published content.

[0161] Based on the above description, this application proposes a cover generation method, which can be executed by the aforementioned electronic device, specifically... Figure 1 The terminal devices and servers shown are illustrated. Please refer to [link / reference]. Figure 10 , Figure 10 This is a flowchart illustrating a cover generation method provided in an embodiment of this application. (See diagram five.) Figure 10 As shown, the cover generation method of this application embodiment may include the following steps: S501. The server constructs a cover strategy library, and for each category in the cover strategy library, it obtains the related published content with a click-through rate higher than the preset click-through rate threshold, performs cover analysis on the content cover of the related published content for each category, and summarizes at least one cover generation strategy related to each category.

[0162] S502, The terminal device displays a content publishing page for publishing the target video, responds to the trigger operation for the cover generation entry, generates a cover generation request for the target video, and sends the cover generation request to the server.

[0163] S503. The server performs video analysis on the target video to obtain the target video's target classification, video summary information, and frame attribute information for each video frame in the target video.

[0164] S504. The server obtains at least one cover generation strategy related to the target classification from the cover strategy library, and determines at least one cover generation strategy associated with the video summary information from the at least one cover generation strategy related to the target classification. The server then selects target video frames from the video frames of the target video whose frame attribute information matches each associated cover generation strategy.

[0165] S505. When the server determines the image processing strategies for the target video frame based on the associated cover generation strategies, it inputs the target video frame and the image processing prompts constructed for each image processing strategy into the image generation model. The image generation model then performs image processing on the target video frame based on each image processing prompt to obtain at least one candidate cover.

[0166] S506, The server returns at least one candidate cover to the terminal device.

[0167] S507. The terminal device displays at least one candidate cover generated for the target video on the content publishing page.

[0168] S508, the terminal device responds to the selection operation for at least one candidate cover and determines the selected candidate cover as the target cover.

[0169] S509. In response to the content publishing operation on the content publishing page, the terminal device generates a content publishing request containing the target video and the target cover, and sends the content publishing request to the server.

[0170] S510: The server publishes target content that includes the target video and whose cover image is the target video cover image.

[0171] In this embodiment, a cover strategy library can be pre-built. When the terminal device triggers the cover generation entry on the content publishing page, the server analyzes the target video to obtain the target video's target category, video summary information, and frame attribute information of each video frame in the target video. It then retrieves cover generation strategies related to the target category from the cover strategy library and selects a specified number of cover generation strategies suitable for the target video. Based on these strategies, corresponding candidate covers are generated, allowing users to select a specific candidate cover as the target cover. This allows the published content to use the target cover as the published content cover. The cover generation strategy is derived by analyzing the covers of related published content belonging to the target category. If the click-through rate of the related published content covers is higher than a preset click-through rate threshold, a cover generation strategy can be pre-analyzed based on a large number of high-quality content covers under the target category to obtain content covers that can attract online users within that target category. Using this strategy for intelligent cover generation results in high-quality covers that appeal to the public and fit the video content. This more intelligent cover determination method improves the quality of cover generation, making the determined covers attractive and representative, meeting the needs of the video content and style, thereby improving the video's click-through conversion rate.

[0172] Please see Figure 11 , Figure 11 This is a schematic diagram of a cover generation device provided in an embodiment of this application. It should be noted that... Figure 11 The cover generation apparatus shown is used to execute this application. Figure 3 , Figure 4 , Figure 6 , Figure 8 , Figure 10 The methods in the illustrated embodiments are shown only in the parts relevant to the embodiments of this application for ease of explanation; specific technical details are not disclosed. Reference to this application is required. Figure 3 , Figure 4 , Figure 6 , Figure 8 , Figure 10 The example shown. Figure 11 The illustrated cover generation device 1000 may include: a content publishing module 1001 and a cover generation module 1002. Wherein: Content publishing module 1001 is used to display the content publishing page for publishing the target video; the content publishing page includes a cover generation entry point; The cover generation module 1002 is used to respond to the trigger operation of the cover generation entry and display the target cover generated according to the cover generation strategy related to the target category of the target video; the content publishing page is used to publish target content that contains the target video and whose content cover is the target cover; The target cover is obtained by image processing of the target video frame in the target video through a cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of related published content is higher than the preset click-through rate threshold.

[0173] Specifically, the cover generation module 1002, in response to a trigger operation on the cover generation entry point, displays the target cover generated according to the cover generation strategy related to the target video's target category. In response to a trigger operation on the cover generation entry, at least one candidate cover generated for the target video is displayed; each candidate cover is generated according to a cover generation strategy, and the cover generation strategy corresponding to each candidate cover is different; In response to a selection operation for at least one candidate cover, the selected candidate cover is determined as the target cover.

[0174] The cover generation module 1002 is also used for: In response to an edit trigger action targeting the target cover, display the cover editing page for the target cover; In response to the prompt word entry operation on the cover editing page, the edited target cover is displayed. The edited target cover determined by the cover editing page is displayed on the content publishing page. The edited target cover is obtained by image processing of the target cover through the editing prompt word corresponding to the prompt word entry operation.

[0175] The content publishing page displays a details page preview control and a cover preview control; The cover generation module 1002 is also used for: When the preview control on the details page is selected, the content details page in preview state is displayed on the content publishing page, and the target video is displayed on the content details page; When the cover preview control is selected, a content recommendation page in preview state is displayed on the content publishing page; the content recommendation page is used to display content cards, which display the content cover determined by the content publishing page.

[0176] Specifically, the cover generation module 1002, in response to a trigger operation on the cover generation entry point, displays the target cover generated according to the cover generation strategy related to the target video's target category. In response to a video frame selection operation for a target video, the video frame selected in the target video is determined as the selected video frame; In response to a trigger operation targeting the cover generation entry, after processing the selected video frames according to the cover generation strategy related to the target video's target category to obtain the target cover, the target cover is displayed.

[0177] The content publishing page includes a cover selection entry. When the cover selection entry is selected, a cover selection area is displayed on the content publishing page. The cover selection area is used to display one or more target covers and / or one or more uploaded covers uploaded by the target account. The target cover is a cover selected from at least one candidate cover generated for the target video. A candidate cover is generated according to a cover generation strategy, and the cover generation strategy corresponding to each candidate cover is different. The cover generation module 1002 is also used for: In response to the content publishing operation on the content publishing page, determine the multiple covers displayed in the cover multi-select area as reference covers, and publish the target publishing content that includes the target video and whose content cover is the reference cover; When the target content is pushed to the content recommendation page, the content cover displayed is selected from the reference covers according to the cover selection strategy. The cover selection strategy includes at least one of the following: random selection, selection based on the account information of the pushing account.

[0178] The cover generation module 1002 is also used for: Displays the content publishing page for publishing the target text; the content publishing page includes a cover page generation entry point; In response to a trigger operation on the cover generation entry, the target cover generated according to the cover generation strategy related to the target text's target category is displayed; the content publishing page is used to publish content that contains the target text and whose content cover is the target cover; The target cover is obtained by generating an image of the target topic of the target text through a cover generation strategy.

[0179] The cover generation module 1002 is also used for: Displays a content publishing page for publishing target text and at least one target image; the content publishing page includes a cover generation entry; In response to a trigger operation on the cover generation entry, the target cover generated according to the cover generation strategy related to the target text's target category is displayed; the content publishing page is used to publish content that contains the target text and whose content cover is the target cover; The target cover is obtained by image processing of at least one matching image in the target image through a cover generation strategy. The matching image is the image that matches the cover generation strategy.

[0180] In response to a trigger operation on the cover generation entry, the target cover generated according to the cover generation strategy related to the target video's target category is displayed, including: In response to a trigger operation targeting the cover generation entry, generate a cover generation request for the target video; The cover generation request is sent to the server so that the server can perform video analysis on the target video to obtain the target classification, video summary information and frame attribute information of each video frame in the target video. From at least one cover generation strategy related to the target classification, the cover generation strategy associated with the video summary information is determined. From the video frames of the target video, the target video frames whose frame attribute information matches the associated cover generation strategy are selected. When determining the image processing strategy for the target video frame based on the associated cover generation strategy, the target video frame and the image processing prompt words constructed for the image processing strategy are input into the image generation model. The image generation model performs image processing on the target video frame based on the image processing prompt words to obtain the target cover. Upon receiving the target cover returned by the server, display the target cover on the content publishing page.

[0181] Among them, at least one cover generation strategy related to the target category is obtained by clustering the content covers of related published content belonging to the target category and performing cover analysis on the content covers based on the clustering results. One cover generation strategy corresponds to one clustering result, and any clustering result includes at least one content cover. Content cover refers to a cover with a click-through rate (CTR) higher than a preset CTR threshold. Related published content refers to published content belonging to the target category with a CTR higher than a preset CTR threshold.

[0182] The specific implementation methods of the content publishing module and the cover generation module can be found in the description of the above embodiments, and will not be repeated here. It should be understood that the beneficial effects obtained by using the same method will also not be repeated here.

[0183] Please see Figure 12 , Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 12 As shown, the electronic device 2600 includes at least one processor 2601 and a memory 2602. Optionally, the electronic device may also include a network interface. The processor 2601, memory 2602, and network interface can exchange data. The network interface, controlled by the processor 2601, is used to send and receive messages. The memory 2602 stores computer programs, including program instructions. The processor 2601 executes the program instructions stored in the memory 2602. The processor 2601 is configured to invoke the program instructions to execute the aforementioned method. The memory 2602 may include volatile memory, such as random-access memory (RAM); the memory 2602 may also include non-volatile memory, such as flash memory, solid-state drive (SSD), etc.; the memory 2602 may also include combinations of the above types of memory.

[0184] Processor 2601 may be a central processing unit (CPU). In one embodiment, processor 2601 may also be a graphics processing unit (GPU). Processor 2601 may also be a combination of a CPU and a GPU. Processor 2601 may be used to invoke device control applications stored in memory 2602 to perform the above-described tasks. Figure 3 , Figure 4 , Figure 6 , Figure 8 , Figure 10 The description of the cover generation method in the corresponding embodiment can also be executed as described above. Figure 11 The description of the cover generation apparatus in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated.

[0185] In specific implementations, the devices, processors, memory, etc., described in the embodiments of this application can execute the implementation methods described in the above method embodiments, or they can execute the implementation methods described in the embodiments of this application, which will not be repeated here.

[0186] This application also provides a computer-readable storage medium storing a computer program. The computer program includes program instructions, which, when executed by a processor, enable the processor to perform some or all of the steps described in the above method embodiments. Optionally, the computer storage medium can be volatile or non-volatile. The computer-readable storage medium may primarily include a program storage area and a data storage area. The program storage area may store an operating system, at least one application program required for a given function, etc.; the data storage area may store data created based on the use of blockchain nodes, etc.

[0187] This application provides a computer program product, which may include a computer program. When the computer program is executed by a processor, it can implement some or all of the steps in the above method, which will not be elaborated here.

[0188] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0189] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer storage medium, which can be a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the methods described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0190] The above-disclosed embodiments are merely some of the embodiments of this application, and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments, and equivalent changes made in accordance with the claims of this application, still fall within the scope of this application.< / category>

Claims

1. A method for generating a cover, characterized in that, The method includes: Displays a content publishing page for publishing the target video; the content publishing page includes a cover image generation entry point; In response to a trigger operation on the cover generation entry, a target cover generated according to the cover generation strategy related to the target category of the target video is displayed; the content publishing page is used to publish target content that includes the target video and whose content cover is the target cover; The target cover is obtained by image processing of the target video frame in the target video through the cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of the related published content is higher than a preset click-through rate threshold.

2. The method according to claim 1, characterized in that, The response to the trigger operation for the cover generation entry point, displaying the target cover generated according to the cover generation strategy related to the target video's target category, includes: In response to a trigger operation on the cover generation entry, at least one candidate cover generated for the target video is displayed; each candidate cover is generated according to a cover generation strategy, and the cover generation strategy corresponding to each candidate cover is different; In response to the selection operation for the at least one candidate cover, the selected candidate cover is determined to be the target cover.

3. The method according to claim 1, characterized in that, The method further includes: In response to an edit trigger operation on the target cover, a cover editing page for the target cover is displayed; In response to the prompt word input operation on the cover editing page, the edited target cover is displayed, and the edited target cover determined by the cover editing page is displayed on the content publishing page; the edited target cover is obtained by image processing of the target cover through the editing prompt word corresponding to the prompt word input operation.

4. The method according to claim 1, characterized in that, The content publishing page displays a details page preview control and a cover preview control; The method further includes: When the preview control on the details page is selected, the content details page in preview state is displayed on the content publishing page, and the target video is displayed on the content details page; When the cover preview control is selected, a content recommendation page in preview state is displayed on the content publishing page; the content recommendation page is used to display content cards, and the content cards display the content cover determined by the content publishing page.

5. The method according to claim 1, characterized in that, The response to the trigger operation for the cover generation entry point, displaying the target cover generated according to the cover generation strategy related to the target video's target category, includes: In response to a video frame selection operation for the target video, the video frame selected in the target video is determined to be the selected video frame; In response to a trigger operation on the cover generation entry, after performing image processing on the selected video frame according to the cover generation strategy related to the target video's target category to obtain the target cover, the target cover is displayed.

6. The method according to claim 1, characterized in that, The content publishing page includes a cover selection entry. When the cover selection entry is selected, a cover selection area is displayed on the content publishing page. The cover selection area is used to display one or more target covers and / or one or more uploaded covers uploaded by the target account. The target cover is a cover selected from at least one candidate cover generated for the target video. A candidate cover is generated according to a cover generation strategy, and the cover generation strategy corresponding to each candidate cover is different. The method further includes: In response to a content publishing operation on the content publishing page, the multiple covers displayed in the cover multi-select area are determined as reference covers, and target publishing content containing the target video and whose content cover is the reference cover is published; When the target content is pushed to the content recommendation page, the displayed content cover is selected from the reference covers according to the cover selection strategy, which includes at least one of the following: random selection, selection based on the account information of the pushing account.

7. The method according to claim 1, characterized in that, The method further includes: Displays a content publishing page for publishing target text; the content publishing page includes a cover page generation entry point; In response to a trigger operation on the cover generation entry, a target cover generated according to a cover generation strategy related to the target category of the target text is displayed; the content publishing page is used to publish content that contains the target text and whose content cover is the target cover; The target cover is obtained by generating an image of the target theme of the target text using the cover generation strategy.

8. The method according to claim 1, characterized in that, The method further includes: Displays a content publishing page for publishing target text and at least one target image; the content publishing page includes a cover generation entry point; In response to a trigger operation on the cover generation entry, a target cover generated according to a cover generation strategy related to the target category of the target text is displayed; the content publishing page is used to publish content that contains the target text and whose content cover is the target cover; The target cover is obtained by image processing of a matching image in at least one target image using the cover generation strategy, and the matching image is an image that matches the cover generation strategy.

9. The method according to claim 1, characterized in that, The response to the trigger operation for the cover generation entry point, displaying the target cover generated according to the cover generation strategy related to the target video's target category, includes: In response to a trigger operation on the cover generation entry, a cover generation request for the target video is generated; The cover generation request is sent to the server, which then performs video analysis on the target video to obtain the target video's target classification, video summary information, and frame attribute information for each video frame in the target video. The server determines a cover generation strategy associated with the video summary information from at least one cover generation strategy related to the target classification. It then selects target video frames from the video frames whose frame attribute information matches the associated cover generation strategy. When determining an image processing strategy for the target video frame based on the associated cover generation strategy, the server inputs the target video frame and an image processing prompt word constructed for the image processing strategy into an image generation model. The image generation model then performs image processing on the target video frame based on the image processing prompt word to obtain the target cover. Upon receiving the target cover returned by the server, the target cover is displayed on the content publishing page.

10. The method according to claim 9, characterized in that, At least one cover generation strategy related to the target category is obtained by clustering the content covers of related published content belonging to the target category and performing cover analysis on the content covers based on the clustering results. One cover generation strategy corresponds to one clustering result, and any clustering result includes at least one content cover. The content cover refers to a cover with a click-through rate higher than a preset click-through rate threshold, and the associated published content refers to published content belonging to the target category with a click-through rate higher than a preset click-through rate threshold.

11. A cover generating device, characterized in that, The device includes: The content publishing module is used to display a content publishing page for publishing target videos; the content publishing page includes a cover generation entry point; The cover generation module is used to respond to a trigger operation on the cover generation entry and display the target cover generated according to the cover generation strategy related to the target category of the target video; the content publishing page is used to publish target content that includes the target video and whose content cover is the target cover; The target cover is obtained by image processing of the target video frame in the target video through the cover generation strategy. The target video frame refers to the video frame that matches the cover generation strategy. The cover generation strategy is summarized by analyzing the content covers of related published content belonging to the target category. The click-through rate of the content covers of the related published content is higher than a preset click-through rate threshold.

12. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to perform the method as described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-10.

14. A computer program product, characterized in that, The method includes a computer program comprising program instructions that, when executed by a processor, implement the method according to any one of claims 1-10.

Images