Poster generation method and device, storage medium and electronic equipment
By extracting text from template posters and using literary and artistic image models to generate background images, the complexity and efficiency problems in the traditional poster design process are solved, and the intelligent generation and efficient production of posters are realized, which improves the innovation of the design and flexibility.
Patent Information
- Application Number
- CN202411213593.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-06
AI Technical Summary
There are problems in the traditional poster design process, which is complicated in the production process and low design efficiency, making it difficult to meet the marketing poster design needs of large demand, high timeliness and obvious customized features.
By obtaining template posters, extracting scene description text and content label text, forming template prompt text, and using literary picture mockup models to generate background pictures, and finally generating target posters based on background pictures, simplifying the design process and improving efficiency.
It realizes intelligent generation of posters, saves time and labor costs, improves the innovation and flexibility of design, and can complete the production of target posters efficiently and with high quality.
Smart Images

Figure CN119091004A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and in particular to a poster generation method, device, storage medium and electronic device. Background Art
[0002] Marketing posters are a common promotional method adopted by many industries when conducting online and offline marketing activities. They have significant requirements such as large demand, high production timeliness, obvious customization, and diverse styles. Taking the insurance industry as an example, when a new insurance product is launched, it is usually necessary to customize the design and production of marketing posters based on product characteristics, target audiences or marketing activities. The design and production mainly includes background image production and artistic font design.
[0003] At present, the traditional method is for staff to use design tools such as Adobe Photoshop (PS) to design posters. However, this method often has pain points such as complicated production process and low design efficiency during the design and production process. Summary of the invention
[0004] In view of this, the present application provides a poster generation method, device, storage medium and electronic device, the main purpose of which is to improve the technical problems of complex poster production process and low design efficiency in the current prior art.
[0005] In a first aspect, the present application provides a poster generation method, comprising:
[0006] Get the template poster;
[0007] Extracting scene description text and content label text from the template poster to form template prompt text;
[0008] Generate a background image using a large Wenshengtu model according to the template prompt text;
[0009] Based on the background image, a target poster is generated.
[0010] Optionally, extracting scene description text and content label text from the template poster to form template prompt text includes:
[0011] extracting the scene description text from the template poster using a scene recognition model; and,
[0012] Extracting the content label text from the template poster using a masked convolutional neural network model;
[0013] The scene description text and the content label text are combined to obtain the template prompt text.
[0014] Optionally, the extracting the content label text from the template poster using a masked convolutional neural network model includes:
[0015] The image content of the template poster is extracted through a masked convolutional neural network model, and based on the image content, the template poster is instance segmented to generate the content label text.
[0016] Optionally, the training process of the large model of the Wensheng graph includes:
[0017] Construct a poster-style image and text dataset;
[0018] Based on the image and text dataset, an initial text-image model is obtained by training;
[0019] The initial Wensheng graph large model is fine-tuned using LoRA technology to obtain the Wensheng graph large model.
[0020] Optionally, the image and text dataset for constructing a poster-style image includes:
[0021] Get a picture of the poster style;
[0022] Using the BLIP model to perform text annotation on the poster-style picture, and generating descriptive text corresponding to the poster-style picture;
[0023] The image and text dataset is obtained based on the poster-style picture and the descriptive text.
[0024] Optionally, generating a target poster based on the background image includes:
[0025] The background image and the text to be added are input into the image-text fusion model to generate the target poster.
[0026] Optionally, generating a background image using a large model of text and graphics according to the template prompt text includes:
[0027] Preprocessing the template prompt text according to the user's required prompt text;
[0028] The prompt text obtained after preprocessing is input into the Wenshengtu large model to generate the background image.
[0029] In a second aspect, the present application provides a poster generation device, comprising:
[0030] An acquisition module is configured to acquire a template poster;
[0031] An extraction module is configured to extract scene description text and content label text from the template poster to form a template prompt text;
[0032] A first generating module is configured to generate a background image using a large model of text and image according to the template prompt text;
[0033] The second generating module is configured to generate a target poster based on the background image.
[0034] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in the first aspect is implemented.
[0035] In a fourth aspect, the present application provides an electronic device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the method described in the first aspect when executing the computer program.
[0036] By means of the above technical solution, the present application provides a poster generation method, device, storage medium and electronic device. Specifically, firstly, a template poster is obtained; then, scene description text and content label text are extracted from the template poster to form a template prompt text; then, according to the template prompt text, a background image is generated using a large model of a text image; based on the background image, a target poster is generated. Compared with the current existing technology, by applying the technical solution of the present application, template prompt text is extracted from the template poster, and the large model technology is combined to realize the intelligent generation from the background image to the target poster, which simplifies the poster production process, saves time and labor costs, and can complete the production of the target poster efficiently and with high quality, while also improving the innovation and flexibility of the poster design.
[0037] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0040] Figure 1 A schematic diagram of a process of generating a poster provided in an embodiment of the present application is shown;
[0041] Figure 2 A schematic diagram showing a flow chart of another poster generation method provided in an embodiment of the present application is shown;
[0042] Figure 3 A schematic diagram of a process flow of an application example provided by an embodiment of the present application is shown;
[0043] Figure 4 A schematic structural diagram of a poster generating device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION
[0044] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0045] In order to improve the technical problems of the current existing technology of complex poster production process and low design efficiency. This embodiment provides a poster generation method, such as Figure 1 As shown, the method includes:
[0046] Step 101, obtain a template poster.
[0047] For example, a user may provide at least one poster that can be used as a template to provide template information as a reference for subsequent generation of a target poster.
[0048] Step 102: extract scene description text and content label text from the template poster to form template prompt text.
[0049] For example, the scene description text may be a text description of the overall visual effect of the template poster, and the content label text may include the main part of the poster content, such as people, windows, clouds, etc. The template prompt text is a guiding text framework that can be used when generating a specific type of content, including content information in the template poster.
[0050] Step 103: Generate a background image using the large Wenshengtu model according to the template prompt text.
[0051] In some examples, the Wenshengtu model can generate images based on given text descriptions. By inputting the template prompt text into the Wenshengtu model, a background image corresponding to the template prompt text can be quickly generated, reducing the time and labor costs of finding a suitable background image.
[0052] Step 104: Generate a target poster based on the background image.
[0053] For example, a target poster can be quickly generated based on the background image and specific needs. For example, text, icons, graphics and other elements can be added to the background image according to the theme of the poster, and the content, style and details of the image can be adjusted as needed. The generated image can be modified at any time to adapt to changing needs.
[0054] In some examples, the target poster can be generated based on the poster production method of image processing: provide a background image and target text, and use traditional image segmentation and texture synthesis technology to embed the text in the best position of the background image; or the target poster can be generated based on deep learning poster production: provide a target poster, use the image classification model to determine the category style of the poster image that the user wants to generate and the number of required text boxes, and use the deep learning model to select the appropriate image from a batch of preset poster images according to the above categories and the number of text boxes, and finally use image processing technology to fill the text into the image. However, the posters produced in this way are limited to the templates in the database, the posters are less innovative, and a large number of poster templates need to be preset in advance, and the selected templates need to be further edited and modified afterwards, and the production process is complicated and not flexible enough. It is also possible to generate a target poster based on a poster production method based on a large model: this type of solution uses an existing general large model to generate a prompt text for generating an image using a preset poster template, and then inputs the obtained prompt text into the poster intelligent generation model for processing to obtain a poster image. However, this method has the following problems: First, the general large model is used for image generation, and the style type of the generated result is uncontrollable, which is difficult to meet user requirements; second, the existing large model of Wenshengtu generates Chinese characters poorly, and this method of generating posters does not provide a reasonable solution to intelligently and flexibly embed text into images. Traditional poster production methods rely on a series of preset templates, which limits the diversity and innovation of the design. In contrast, this embodiment uses large model technology to intelligently generate background images and artistic characters for posters, which is more creative and flexible.
[0055] This embodiment first obtains a template poster; then extracts scene description text and content label text from the template poster to form a template prompt text; then generates a background image based on the template prompt text using the Wenshengtu large model; and generates a target poster based on the background image. Compared with the current existing technology, this embodiment extracts the template prompt text from the template poster, and combines the large model technology to realize the intelligent generation from the background image to the target poster, which simplifies the poster production process, saves time and labor costs, and can complete the production of the target poster efficiently and with high quality, while also improving the innovation and flexibility of poster design.
[0056] Further, as a refinement and extension of the above embodiment, in order to fully illustrate the specific implementation process of the method of this embodiment, this embodiment provides the following Figure 2The specific method shown includes:
[0057] Step 201: Get a template poster.
[0058] In some examples, such as Figure 3 As shown, if the user cannot provide a poster background image template, the user requirement prompt text can be directly pre-processed, and then the text image model can be input to generate the background image.
[0059] Step 202: extract scene description text from the template poster using a scene recognition model; and extract content label text from the template poster using a masked convolutional neural network model.
[0060] For example, Figure 3 As shown in the figure, if the user can provide a template / sample of the marketing poster background image, it can be input into the scene recognition model BLIP (Bootstrapping Language-Image Pre-training) and the region-based mask convolutional neural network model Mask R-CNN (Region-based Convolutional Neural Network with Mask) at the same time. Among them, the BLIP model is a multimodal pre-training model that can understand the content of the image and generate descriptive text to achieve image generation. The Mask R-CNN model can be used for object detection and instance segmentation, that is, to identify each object instance in the image and generate an accurate pixel-level mask for each instance.
[0061] For example, the scene recognition model can be used to extract and output the scene description text of the poster background template, such as: [There is a car and a kitten under the tree]. By using the scene recognition model and the mask convolutional neural network model to extract the poster background image template information, it is possible to generate a poster background image that reflects both the reference template information and personalized user needs.
[0062] Optionally, step 202 may specifically include: extracting the image content of the template poster through a masked convolutional neural network model, and performing instance segmentation on the template poster based on the image content to generate content label text.
[0063] For example, the template poster image can be read first and properly preprocessed, such as scaling and normalization, and then the pre-trained Mask R-CNN model can be used to detect different instances in the image and generate a segmentation mask for each instance. Then, based on the segmentation mask, features are extracted for each instance to generate content label text.
[0064] For example, the Mask R-CNN model can be used to extract the main image content of a poster background template, and perform instance segmentation on the poster background image to generate one or more content label texts, such as: [kitten, car, big tree].
[0065] Step 203: Combine the scene description text and the content label text to obtain a template prompt text.
[0066] For example, the scene description text [There is a car and a kitten under the big tree] output by the BLIP model and the content label text [kitten, car, big tree] output by the Mask R-CNN model can be combined to form a template prompt text, such as: [There is a car and a kitten under the big tree, kitten, car, big tree].
[0067] In some examples, the scene description text provides information about the image environment and is a comprehensive summary of the entire image content, including detailed descriptions and contextual information, while the content label text clarifies the specific elements in the image. For complex scenes in an image, it may be difficult to fully express the meaning of the image by relying solely on the content label text. The scene description text can provide background information to help people better understand the content expressed by the image. Combining the two can provide richer and more detailed information for the template image, improve the accuracy and fluency of the template prompt text, reduce ambiguity, and enable the subsequent generation of more accurate and comprehensive background images.
[0068] Step 204: Generate a background image using the large Wenshengtu model according to the template prompt text.
[0069] In some examples, the Wenshengtu model is a large-scale machine learning model that uses deep learning methods to generate images based on text descriptions. A large number of images and corresponding text annotations are usually used for training so that the relationship between text and image can be understood and captured, and then the corresponding image can be generated based on the new text input. Specifically, the input text is first converted into a vector form by an encoder, and the vector contains the semantic information of the text. The encoded text information will be used to guide the generator (usually based on a diffusion model or a generative adversarial network) to generate images. In order to ensure that the generated image is consistent with the text description, a conditional generation method is usually used, that is, text condition constraints are added during the generation process. By using the Wenshengtu model to generate background images, images can be customized according to specific needs, such as color, style, etc., which reduces labor and time costs, can quickly generate images, and has a wide range of application scenarios.
[0070] Optionally, the training process of the large text-image model may specifically include: constructing a picture-text dataset of poster-style pictures; training an initial large text-image model based on the picture-text dataset; and fine-tuning the initial large text-image model using LoRA technology to obtain the large text-image model.
[0071] In some examples, the initial text image model, such as the Stable Diffusion text image model, can be fine-tuned using the LoRA (Low-Rank Adaptation) technique to obtain a text image model specifically for the marketing poster style. LoRA is a fine-tuning technique for deep learning models that can combine tasks of visual and text data, such as image description generation.
[0072] For example, weight matrix decomposition can be achieved by decomposing the weight matrix in the model into the product of two low-rank matrices, and only the decomposed low-rank matrix needs to be updated instead of the weight of the entire model, so that the model can adapt to new tasks or fields more quickly. Combining the parameters fine-tuned by LoRA with the parameters of the base model can achieve better performance.
[0073] For example, the goal of LoRA is to reduce the number of parameters required in the fine-tuning process, thereby saving computing resources. Applying LoRA fine-tuning technology to the production of marketing posters can be customized according to user needs. Training in different styles can be developed to meet user needs, effectively solving the problem of difficult control of poster styles generated by general large models.
[0074] Optionally, the above-mentioned construction of a graphic and text dataset of poster-style pictures may specifically include: obtaining poster-style pictures; using a BLIP model to perform text annotation on the poster-style pictures to generate descriptive text corresponding to the poster-style pictures; and obtaining a graphic and text dataset based on the poster-style pictures and the descriptive text.
[0075] Exemplarily, the image data in the image and text dataset can be obtained from the Internet and / or a business party, and the text is annotated through the open source model BLIP, and then manually modified.
[0076] Optionally, step 204 may specifically include: preprocessing the template prompt text according to the prompt text required by the user; inputting the prompt text obtained after the preprocessing into the text image model to generate a background image.
[0077] For example, the personalized needs of users for the background image of the marketing poster to be generated can be obtained first, including the image size, image subject, etc., to form a user demand prompt text. For example, the user demand prompt text can be [50cm×70cm, car, gift box], etc. The user demand prompt text can be used to clarify the user's needs, understand the effect the user wants to achieve, and generate a background image that adapts to the user demand prompt text.
[0078] Exemplarily, the template prompt text and the user demand prompt text can be input into the prompt text preprocessing module, through which the template prompt text and the user demand prompt text are preprocessed, that is, the text content is deduplicated and spliced. For example, if the above template prompt text and the demand prompt text both contain [car], deduplication processing can be performed, and the spliced prompt text is: [50cm×70cm, there is a car and a kitten under the big tree, car, gift box, kitten, big tree]. The prompt text obtained after preprocessing is input into the trained special marketing poster style Wenshengtu large model for processing, and the poster background image can be generated.
[0079] Step 205: Generate a target poster based on the background image.
[0080] Optionally, step 205 may specifically include: inputting the background image and text to be added into the image-text fusion model to generate a target poster.
[0081] In some examples, the image-text fusion model refers to an artificial intelligence model that can process image and text data simultaneously, and can be used to perform tasks such as image description generation, visual question answering, text-to-image synthesis, cross-modal retrieval, and dialogue systems. The features of the image and text are extracted through a multimodal encoder and converted into a unified representation space. The image-text fusion model can be trained using image and text data at the same time, so that the model can learn the association between cross-modalities, and can also use pre-trained image recognition models and language models as a basis for fine-tuning specific tasks.
[0082] For example, based on the generated background image, you can input the text you want to add into the image-text fusion model, such as "Congratulatory message" or "Signing an order of XX million", and edit the location of the added text (you can drag the frame on the background image), so that text information can be generated at the specified location of the image, completing the process of making marketing posters from images to artistic words. Using the image-text fusion model to generate artistic words can make the fusion of artistic words and background images more natural and harmonious.
[0083] The method proposed in the present disclosure is applied to the generation of target posters, and its application has a variety of scenarios. The following are some possible application fields and scenarios. In the poster production scenario: when the user can provide a template poster as a reference sample, the scene description information and the content body information can be extracted from the background image of the template poster, wherein the scene description information and the content body information can be in the form of text. For example, the scene description information is [There is a desk lamp and a water cup on the desk], and the content body information is [Desk, desk lamp, water cup]. Combining the two types of information, the template prompt text can be obtained: [There is a desk lamp and a water cup on the desk, desk, desk lamp, water cup]. Then the above template prompt text is input into the Wenshengtu large model dedicated to the marketing poster template, and the background image that meets the description of the template prompt text can be obtained. The information in the background image corresponds to [There is a desk lamp and a water cup on the desk, desk, desk lamp, water cup]. The above background image and the text to be added and the location information of the text are input into the picture-text fusion large model to generate the target poster. In product marketing and promotion scenarios, posters can be intelligently generated to promote the features, advantages, and launch dates of new products, meeting user requirements for style and promotional direction. Natural and harmonious artistic wording can better attract the attention of target customers.
[0084] Compared with the current existing technology, this embodiment realizes the intelligent and process-based production of target posters by extracting template prompt text from template posters and combining multiple large model technologies to complete the generation of target posters. Compared with the cumbersome process of traditional methods, it saves time and labor costs, can complete the production of target posters with high efficiency and high quality, and also improves the innovation and flexibility of poster design. It can generate posters of different styles according to the personalized needs of users, and enhance the user experience.
[0085] Further, as Figure 1 The specific implementation of the method shown in the embodiment provides a poster generation device, such as Figure 4 As shown, the device includes: an acquisition module 31, an extraction module 32, a first generation module 33, and a second generation module 34.
[0086] An acquisition module 31 is configured to acquire a template poster;
[0087] An extraction module 32 is configured to extract scene description text and content label text from the template poster to form a template prompt text;
[0088] The first generating module 33 is configured to generate a background image using a large model of a text image according to the template prompt text;
[0089] The second generating module 34 is configured to generate a target poster based on the background image.
[0090] In some examples, the extraction module 32 is specifically configured to extract the scene description text from the template poster using a scene recognition model; and to extract the content label text from the template poster using a masked convolutional neural network model; and to combine the scene description text and the content label text to obtain the template prompt text.
[0091] In some examples, the extraction module 32 is further configured to extract the image content of the template poster through a masked convolutional neural network model, and perform instance segmentation on the template poster based on the image content to generate the content label text.
[0092] In some examples, the first generation module 33 is specifically configured to construct a picture and text dataset of poster-style pictures; based on the picture and text dataset, train an initial text-image large model; and use LoRA technology to fine-tune the initial text-image large model to obtain the text-image large model.
[0093] In some examples, the first generation module 33 is further specifically configured to obtain the poster-style picture; use the BLIP model to perform text annotation on the poster-style picture to generate descriptive text corresponding to the poster-style picture; and obtain the image and text dataset based on the poster-style picture and the descriptive text.
[0094] In some examples, the second generation module 34 is specifically configured to input the background image and the text to be added into the image-text fusion model to generate the target poster.
[0095] In some examples, the first generation module 33 is further configured to pre-process the template prompt text according to the user's prompt text requirement; input the prompt text obtained after the pre-processing into the Wenshengtu model to generate the background image.
[0096] Based on the above Figure 1 and Figure 2 The method shown in the embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned Figure 1 and Figure 2 The method shown.
[0097] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of the present application.
[0098] Based on the above Figure 1 and Figure 2 The method shown, and Figure 4 In order to achieve the above-mentioned purpose, the embodiment of the present application further provides an electronic device, which may include a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figure 1 and Figure 2 The method shown.
[0099] Optionally, the above-mentioned physical device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.
[0100] Those skilled in the art will appreciate that the above-mentioned physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or a combination of certain components, or different arrangements of components.
[0101] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the above-mentioned physical device, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing physical device.
[0102] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. By applying the solution of this embodiment, the template prompt text is extracted from the template poster, and the automatic generation from the background image to the artistic text is realized by combining a variety of large model technologies, thereby completing the intelligent production of the target poster. By applying the method of this embodiment, compared with the cumbersome process of the traditional method, time and labor costs are saved, and the production of the target poster can be completed efficiently and with high quality, while also improving the innovation and flexibility of the poster design.
[0103] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprising a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0104] The above is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments herein, but will conform to the widest scope consistent with the principles and novel features applied for herein.
Claims
1. A poster generation method, characterized in that: include: Get the template poster; Extracting scene description text and content label text from the template poster to form template prompt text; Generate a background image using a large Wenshengtu model according to the template prompt text; Based on the background image, a target poster is generated.
2. The method according to claim 1, characterized in that The step of extracting scene description text and content label text from the template poster to form template prompt text includes: extracting the scene description text from the template poster using a scene recognition model; and, Extracting the content label text from the template poster using a masked convolutional neural network model; The scene description text and the content label text are combined to obtain the template prompt text.
3. The method according to claim 2, characterized in that The extracting the content label text from the template poster using a masked convolutional neural network model includes: The image content of the template poster is extracted through a masked convolutional neural network model, and based on the image content, the template poster is instance segmented to generate the content label text.
4. The method according to claim 1, characterized in that: The training process of the Wensheng graph large model includes: Construct a poster-style image and text dataset; Based on the image and text dataset, an initial text-image model is obtained by training; The initial Wensheng graph large model is fine-tuned using LoRA technology to obtain the Wensheng graph large model.
5. The method according to claim 4, characterized in that The image and text dataset for constructing poster-style images includes: Get a picture of the poster style; Using the BLIP model to perform text annotation on the poster-style picture, and generating descriptive text corresponding to the poster-style picture; The image and text dataset is obtained based on the poster-style picture and the descriptive text.
6. The method according to claim 1, characterized in that The step of generating a target poster based on the background image includes: The background image and the text to be added are input into the image-text fusion model to generate the target poster.
7. The method according to claim 1, characterized in that The step of generating a background image using a large model of a text image according to the template prompt text includes: Preprocessing the template prompt text according to the user's required prompt text; The prompt text obtained after preprocessing is input into the Wenshengtu large model to generate the background image.
8. A poster generating device, characterized in that: include: An acquisition module, configured to acquire a template poster; An extraction module is configured to extract scene description text and content label text from the template poster to form a template prompt text; A first generating module is configured to generate a background image using a large model of text and image according to the template prompt text; The second generating module is configured to generate a target poster based on the background image.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
10. An electronic device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Advertisement picture generation method and device, storage medium and program product
CN119762158A
Personalized image generation method and device, electronic equipment and storage medium
CN120612396A
