AI story picture book generation method and device based on parallel processing, and terminal

Abstract

The invention discloses an AI story picture book generation method and device based on parallel processing and a terminal. The method comprises the steps of obtaining a story generation request; calling an AI large model to automatically generate a complete story text according to the story generation request; extracting the first-segment text content of the complete story text, starting an independent thread, and performing AI story picture book generation processing on the first-segment text content of the complete story text; when the AI story picture book of the first text content is generated and processed, the AI story picture book is timely returned to the front end for playing; meanwhile, creating a thread pool to dynamically allocate processing tasks of subsequent paragraphs, and sequentially performing dynamic AI story picture book generation processing on text contents of the paragraphs behind the complete story text; and when the playing of the AI story picture book of the first segment of text content is completed, sequentially playing the AI story picture books of the subsequent segments.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence application technology, and in particular to a method, device, intelligent terminal and storage medium for generating an AI story picture book based on parallel processing. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, AI-generated content (AIGC) has been widely applied in various fields. Among them, AI storybook generation, as an emerging content creation method, is attracting increasing attention from users. Currently, AI storybook generation technology typically includes multiple steps, such as story text generation, text-to-speech conversion, and text-to-image generation. The processing efficiency and user experience of these steps directly affect the practicality of the entire system.

[0003] In the prior art, CN117877052A discloses an interactive story picture book generation method. This method constructs anchor information based on user input, generates a story protagonist and title, and rewrites or continues the story content based on user interaction instructions. CN119922359A proposes a display device and display method that generates a picture book through voice interaction and displays the picture book generation progress and final picture book content on a display. CN117332118A describes a story video generation method that generates story text and predicts story images containing character images based on user-input storyline information and selected character images.

[0004] Furthermore, CN110782900B discloses a collaborative AI storytelling system that provides an impromptu storytelling AI agent that can collaborate and interact with users. The system includes multiple components such as natural language understanding, processing, and generation. CN118428331A proposes an information generation method that generates an initial body text based on preset basic story information and generates the final work text based on user feedback information.

[0005] However, existing AI story picture book generation technology has the following problems: First, most systems use a serial processing method, that is, the entire story text is generated first, and then the audio and image generation of each paragraph is processed in turn, resulting in users having to wait for all content generation to be completed (usually takes 3-5 minutes) before they can start reading. The waiting time on the first screen is too long, which seriously affects the user experience. Secondly, the serial processing method between paragraphs leads to insufficient utilization of computing resources. The CPU / GPU utilization rate usually does not exceed 40%, resulting in idle and wasted computing resources. Third, the existing technology lacks an effective generation progress visualization and exception handling mechanism. Users cannot understand the progress status of content generation in real time. When encountering generation exceptions, there is also a lack of timely feedback, resulting in an incoherent interactive experience.

[0006] The existence of the above problems makes the existing AI story picture book generation system face challenges such as poor user experience, low resource utilization, and unsmooth interaction in actual applications. There is an urgent need for a technical solution that can improve processing efficiency and optimize user experience. Summary of the Invention

[0007] In order to solve the technical problems of long waiting time for the first screen, idle and wasted computing resources, and discontinuous interactive experience, and to achieve the technical effects of reduced latency, improved resource utilization, and optimized QoE, the present invention provides an AI story picture book generation method, device, smart terminal, and storage medium based on parallel processing. The present invention greatly reduces user waiting time, improves user experience, and provides convenience for users.

[0008] The technical solutions adopted by the present invention to solve the problem are as follows: Provided is an AI story picture book generation and processing method based on parallel processing, comprising: obtaining a story generation request; calling an AI large model to automatically generate a complete story text according to the story generation request; extracting the first paragraph of the complete story text, starting an independent thread, and performing AI story picture book generation processing on the first paragraph of the complete story text; when the AI ​​story picture book generation processing of the first paragraph of the text content is completed, promptly returning it to the front end for playback; and simultaneously creating a thread pool to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of subsequent paragraphs of the complete story text in sequence; when the AI ​​story picture book playback of the first paragraph of the text content is completed, the AI ​​story picture books of subsequent paragraphs are played in sequence.

[0009] Preferably, the step of obtaining a story generation request includes: obtaining a story generation request input by a user, and performing sensitive word detection on the story generation request input by the user.

[0010] Preferably, the step of calling the AI ​​big model to automatically generate a complete story text according to the story generation request includes: calling the AI ​​big model to automatically generate a complete story text according to the story generation request; and using a regular expression to represent the generated complete story text: "Paragraph 1: (.*?)(?=\\s*paragraph[2-9]|$)" to match the first paragraph; and establishing a paragraph queue Q={P1,P2,...,Pn} for paragraphs 1 to paragraph n respectively.

[0011] Preferably, the steps of extracting the first paragraph of the text content of the complete story text, starting an independent thread, and performing AI story picture book generation processing on the first paragraph of the text content of the complete story text include: extracting the first paragraph of the text content of the complete story text; and controlling the start of an independent thread T1 to perform AI story picture book generation processing on the first paragraph of the text content P1 of the complete story text; when performing AI story picture book generation processing on the first paragraph of the text content P1, first generate the first paragraph of audio in a specified audio format according to the first paragraph of the text content P1; and perform AI text picture processing on the first paragraph of the text content P1 to generate the first paragraph of the text picture, and establish a correspondence between the first paragraph of the audio and the first paragraph of the text picture.

[0012] Preferably, the step of performing AI text image processing based on the first paragraph text content P1 to generate the first paragraph text image, and establishing a correspondence between the first paragraph audio and the first paragraph text image also includes: uploading the first paragraph audio and the first paragraph text image with the established correspondence to the front-end cloud storage platform, and adding a corresponding watermark to the first paragraph text image.

[0013] Preferably, the steps of: when the AI ​​story picture book generation processing of the first paragraph of text content is completed, returning it to the front end for playback in time; and simultaneously creating a thread pool to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence include: when the AI ​​story picture book generation processing of the first paragraph of text content is completed, obtaining the AI ​​story picture book corresponding to the first paragraph of text content, and promptly sending it to the front end for real-time playback; while the AI ​​story picture book corresponding to the first paragraph of text content is playing, creating a thread pool Pool={T2, T3, ..., Tn} in parallel in the background to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence, processing and generating audio and text images of corresponding paragraphs in sequence, and establishing a corresponding relationship between the corresponding audio and text images.

[0014] Preferably, the step of playing the AI ​​story picture book of the subsequent paragraphs in sequence after the playback of the AI ​​story picture book of the first paragraph of text content is completed includes: when the playback of the AI ​​story picture book of the first paragraph of text content is completed, according to the processing progress pushed in real time, the AI ​​story picture book of the subsequent paragraphs that have been processed are obtained from the background in sequence, and the AI ​​story picture book of the subsequent paragraphs are played in sequence after progressive rendering.

[0015] The present invention achieves a fast response time of ≤12 seconds for the first paragraph by adopting a technical solution of prioritizing the first paragraph and processing subsequent paragraphs in parallel. Compared with the traditional method that requires waiting for the entire content to be generated (3-5 minutes) before starting to read, the user's waiting time is greatly reduced; by creating a thread pool to process subsequent paragraphs in parallel, the CPU occupancy rate is increased by 40%, effectively solving the problem of idle and wasted computing resources caused by serial processing between paragraphs; at the same time, through real-time push processing progress and progressive rendering technology, the user's reading progress is synchronized with the background processing progress, which significantly improves the consistency of the interactive experience and solves the problem of lack of generation progress visualization and exception handling mechanism.

[0016] The present invention also provides an AI story picture book generation device based on parallel processing, wherein the device includes: The acquisition module is used to obtain story generation requests; The text generation module is used to call the AI ​​model to automatically generate the complete story text based on the story generation request; A first paragraph picture book generation module is used to extract the first paragraph text content of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph text content of the complete story text; The first paragraph playback control and parallel processing module is used to promptly return the AI ​​story picture book generation processing of the first paragraph of text content to the front-end for playback; and at the same time, create a thread pool to dynamically allocate the processing tasks of subsequent paragraphs, and dynamically generate AI story picture books for the text content of the subsequent paragraphs of the complete story text in sequence; The subsequent playback control module is used to play the AI ​​story picture book of subsequent paragraphs in sequence after the AI ​​story picture book of the first paragraph of text content is played.

[0017] An intelligent terminal includes a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, including the method for executing any one of the methods described above.

[0018] A computer-readable storage medium, wherein when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform any one of the methods described above.

[0019] The present invention provides a method, device, intelligent terminal, and storage medium for generating AI storybooks based on parallel processing. The present invention provides a low-latency picturebook generation method that reduces user waiting time from minutes to seconds by combining instant response to the first paragraph with multi-threaded parallel processing. Specifically, after the first paragraph is processed, it is returned to the front end. Once the first paragraph is played, subsequent paragraphs are also generated, significantly reducing user waiting time and providing convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 1 is a flow chart of the AI ​​story picture book generation method based on parallel processing provided in Example 1 of the present invention.

[0022] Figure 2 This is a flowchart of the AI ​​story picture book generation method based on parallel processing provided in Example 2 of the present invention.

[0023] Figure 3 A block diagram of the principles of an embodiment of an AI story picture book generation device based on parallel processing provided by the present invention.

[0024] Figure 4 This is a block diagram of the internal structure of the smart terminal provided by an embodiment of the present invention.

[0025] Figure 5 2 is a schematic diagram of the processing process of the AI ​​story picture book generation method based on parallel processing provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clear and distinct, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0028] The existing AI story picture book generation technology has the following problems: First, most systems use a serial processing method, that is, the entire story text is generated first, and then the audio and image generation of each paragraph is processed in turn. As a result, users need to wait for all content generation to be completed (usually takes 3-5 minutes) before they can start reading. The waiting time on the first screen is too long, which seriously affects the user experience. Secondly, the serial processing method between paragraphs leads to insufficient utilization of computing resources. The CPU / GPU utilization rate usually does not exceed 40%, resulting in idle and wasted computing resources. Third, the existing technology lacks effective generation progress visualization and exception handling mechanisms. Users cannot understand the progress status of content generation in real time. When encountering generation exceptions, there is also a lack of timely feedback, resulting in an incoherent interactive experience.

[0029] This invention provides a low-latency AI storybook generation method based on parallel processing. By combining instant response to the first paragraph with multi-threaded parallel processing, it reduces user waiting time from minutes to seconds. Specifically, after the first paragraph is processed, it is returned to the front-end. After the first paragraph is played, subsequent paragraphs are also generated, significantly reducing user waiting time.

[0030] Example 1 like Figure 1 As shown, a method for generating an AI story picture book based on parallel processing in the first embodiment of the present invention includes the following steps: Step S100: Obtain a story generation request; Specifically, users can enter a story generation request through the client application program interface. This request can include content such as the story theme, character settings, and plot requirements. Upon receiving the user's story generation request, the system performs a sensitive word check on the request. This check uses a pre-established sensitive word library to match the user's input to determine if it contains inappropriate content. If a sensitive word is detected, the system prompts the user to modify the request. If no sensitive word is detected, the system proceeds to the next step.

[0031] Step S200: Based on the story generation request, call the AI ​​big model to automatically generate the complete story text; Specifically, the system takes a story generation request that has been checked for sensitive words as input and invokes a pre-trained AI language model, such as GPT-4, Claude, or similar large-scale language models. Based on the input request, the AI ​​model automatically generates a complete story text. The generated story text includes a complete plot, character dialogue, and scene descriptions, and has a reasonable story structure.

[0032] The system processes the generated complete story text using regular expressions, segmenting it into multiple paragraphs. The regular expression can be expressed as: paragraph 1, paragraph 2 through paragraph n, where paragraph 1 matches the first paragraph. After segmentation, the system creates a paragraph queue Q = {P1, P2, ..., Pn} for paragraphs 1 through n, where P1 represents the first paragraph and P2 through Pn represent subsequent paragraphs. This paragraph queue Q is used to allocate subsequent parallel processing tasks.

[0033] Step S300: extract the first paragraph of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph of the complete story text; Specifically, the system extracts the first paragraph of text content P1 from the paragraph queue Q. The system controls the launch of an independent thread T1, which is specifically used to process the AI ​​storybook generation based on the first paragraph of text content P1. In independent thread T1, the system first processes the first paragraph of text content P1 and generates the first audio in a specified audio format. The audio generation uses text-to-speech technology to convert the text content into natural and fluent speech, which is then saved in a specified audio format such as MP3 or WAV.

[0034] The system also performs AI-generated text-image processing based on the first paragraph's text content, P1, to generate the first paragraph's text-image. This text-image processing can utilize AI image generation models such as StableDiffusion or DALL-E to convert the text description into an image that matches the story content. The generated image can have a resolution of 1024×1024 pixels and be saved in PNG format to ensure image quality.

[0035] The system establishes a correspondence between the generated first audio clip and the first text image, forming the complete first paragraph of the AI ​​storybook content. This correspondence is linked via a unique identifier to ensure synchronized playback of the audio and image. The system uploads the corresponding first audio clip and first text image to the front-end cloud storage platform and adds a corresponding watermark to the first text image. The watermark contains the application logo and copyright information and is located in the lower right corner of the image. The transparency is set to 30% to ensure that it does not affect the user's viewing experience.

[0036] Step S400: When the AI ​​story picture book generation process of the first paragraph of text content is completed, it is promptly returned to the front end for playback; and at the same time, a thread pool is created to dynamically allocate processing tasks for subsequent paragraphs, and the AI ​​story picture book generation process is dynamically performed on the text content of the subsequent paragraphs of the complete story text in sequence; Specifically, once the AI ​​storybook generation process for the first paragraph of text content P1 is complete, the system retrieves the AI ​​storybook corresponding to the first paragraph (including audio and images) and promptly sends it to the front-end application via the network interface for real-time playback. Upon receiving the first AI storybook, the front-end application immediately begins playback, providing instant feedback to the user.

[0037] While the first AI storybook segment plays, the system in this embodiment of the present invention creates a thread pool in parallel in the background to dynamically allocate processing tasks for subsequent segments. The thread pool size is dynamically adjusted based on server resources; for example, the initial setting may be four worker threads. The system sequentially extracts the content of segments P2 through Pn from the segment queue Q and assigns them to idle threads in the thread pool for processing.

[0038] After receiving the paragraph content, each worker thread executes the same process as the first paragraph: first, generating the corresponding paragraph's audio, then generating the corresponding text image, and finally establishing a correspondence between the audio and text image. The processed AI storybook paragraph content is cached on the server, awaiting front-end requests. The system records the processing progress of each paragraph in real time and pushes processing status information to the front-end through technologies such as WebSocket.

[0039] Step S500: When the AI ​​story picture book of the first paragraph of text content is played, the AI ​​story picture book of the subsequent paragraphs are played in sequence.

[0040] Specifically, once the front-end finishes playing the first AI storybook segment, the system retrieves the processed subsequent segments from the back-end based on real-time progress information. The system uses progressive rendering technology to preload the next segment while playing the current segment, ensuring continuous and smooth playback.

[0041] The front-end application plays the AI ​​storybook's subsequent paragraphs in the order P2, P3, ..., Pn. Each paragraph's playback involves synchronously displaying the image and playing the corresponding audio. When switching between paragraphs, the system applies a fade-in and fade-out effect to make the transition more natural. If processing of a paragraph is not complete, the system displays a loading animation until the content of that paragraph is ready.

[0042] Throughout the playback process, the system continuously monitors network status and resource usage, adjusting loading strategies as needed to ensure a smooth user experience. Once all segments have finished playing, the system provides the option to replay or generate a new story, enhancing the user's interactive experience.

[0043] In this way, the present invention reduces user waiting time from minutes to seconds by combining instant response to the first paragraph with multi-threaded parallel processing. That is, after the first paragraph is processed, it is returned to the front end. After the first paragraph is played, the subsequent paragraphs are also generated, which greatly reduces user waiting time.

[0044] It has the following advantages: 1) The waiting time for the first screen is greatly shortened: users do not need to wait for the entire content to be generated (3-5 minutes), they can read as long as the first paragraph of content is generated; 2) Computing resources can be fully utilized: Parallel processing is used between paragraphs, greatly improving CPU / GPU utilization; 3) The interactive experience is more coherent: it has a generation progress visualization and exception handling mechanism.

[0045] Example 2 A second embodiment of the present invention provides an AI story picture book generation method based on parallel processing, comprising the following steps: Step 1: Get a story generation request; Specifically, the system receives a story generation request from a user via a mobile app or web interface. Users can enter their request via a text input box or submit it via voice input-to-text conversion. Upon receiving the story generation request, the system immediately initiates the sensitive word detection process.

[0046] Sensitive word detection utilizes a two-tiered filtering mechanism: initial screening using a local sensitive word library, followed by in-depth detection via a cloud-based API. The sensitive word library contains words from various designated sensitive categories and is regularly updated to maintain its effectiveness. If sensitive content is detected, the system will flag the specific sensitive word and prompt the user to modify it. If the content passes detection, the system proceeds to the next step.

[0047] Step 2: Based on the story generation request, call the AI ​​big model to automatically generate the complete story text; Specifically, the system encapsulates story generation requests that have been tested for sensitive words into a standard API request format, including parameters such as the request content, target story length, and style preferences. Based on the current load, the system selects the most suitable model instance from the large AI model service pool, such as GPT-4, Claude2, or other language models optimized specifically for children's stories.

[0048] After receiving the request, the AI ​​model automatically generates the complete story text based on its built-in story generation algorithm. During the generation process, the model considers factors such as the story's structural integrity, plot coherence, and character development to ensure the generated story is suitable for the target audience. After generation is complete, the system conducts a preliminary quality assessment of the story text to ensure that the content meets the expected requirements.

[0049] The system uses regular expressions to structure the generated complete story text. Regular expression patterns are designed to identify paragraph delimiters (such as two consecutive line breaks) or paragraph start markers, segmenting the text into multiple independent paragraphs. The segmented paragraphs are numbered sequentially, forming a sequence of paragraphs 1, 2, and n, with paragraph 1 being marked as the first paragraph.

[0050] The system creates a paragraph queue Q = {P1, P2, ..., Pn} for each segmented paragraph. The queue uses a first-in, first-out (FIFO) data structure to facilitate sequential processing. Each element in the queue contains information such as the paragraph text, paragraph ID, and processing status, providing the data foundation for parallel processing.

[0051] Step 3: Extract the first paragraph of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph of the complete story text; Specifically, the system extracts the first paragraph of text P1 from the paragraph queue Q and marks it as a priority. The system then starts an independent thread T1 and allocates high-priority resources to it to ensure that the first paragraph can be processed quickly.

[0052] In independent thread T1, the system first performs semantic analysis on the first paragraph of text, P1, to extract key scenes, characters, and emotional tone. Based on this analysis, the system invokes the text-to-speech engine to convert the first paragraph into natural, fluent speech. During speech generation, the system adjusts the tone, speed, and emotional expression based on the text content, making the generated audio more expressive. The generated audio uses a 16-bit / 44.1kHz sampling rate and is saved in MP3 format, ensuring clear sound quality while minimizing file size.

[0053] At the same time, the system performs AI text-generated image processing based on the first paragraph of text content P1. The system first extracts visual keywords and scene descriptions from the text and constructs detailed image generation prompts. The prompt content includes elements such as scene descriptions, character features, and color styles to guide the AI ​​image generation model to create images that are highly matched with the text content. The system calls AI image generation models such as StableDiffusion or DALL-E and uses the constructed prompts to generate the first paragraph of text-generated image. The generated image uses a resolution of 1024×1024 pixels and a color depth of 24 bits and is saved in PNG format.

[0054] The system establishes a correspondence between the first generated audio clip and the first generated text image, creating a metadata file containing the link between the two. This metadata file, formatted in JSON, records information such as the audio file path, image file path, timestamp, and segment ID. The system then uploads the corresponding first audio clip, first generated text image, and metadata file to the front-end cloud storage platform. The upload process utilizes block-by-block upload technology to ensure reliable transmission of large files.

[0055] After the upload is complete, the system adds a corresponding watermark to the first paragraph of the text image. The watermark content includes the application logo, copyright information, and a unique identification code. The watermark is located in the lower right corner of the image and uses a semi-transparent effect with a transparency of 30%, ensuring that it does not affect the user viewing experience while protecting intellectual property rights.

[0056] Step 4: When the AI ​​story picture book generation processing of the first paragraph of text content is completed, it is promptly returned to the front-end for playback; and at the same time, a thread pool is created to dynamically allocate the processing tasks of subsequent paragraphs, and the AI ​​story picture book generation processing is dynamically performed on the text content of the subsequent paragraphs of the complete story text in sequence; Specifically, once the AI ​​storybook generation process for the first paragraph of text content P1 is complete, the system pushes a notification of completion to the front-end via WebSocket or HTTP long link, along with a resource access link. Upon receiving the notification, the front-end application immediately retrieves the first paragraph of the AI ​​storybook content (including audio, images, and metadata) from the cloud storage platform and begins playback.

[0057] While the first AI storybook plays, the system creates a dynamic thread pool in the background. The initial size of the thread pool is determined by the number of CPU cores and memory available on the server, typically 4 to 8 worker threads. The thread pool uses an adaptive adjustment strategy, dynamically adjusting the number of threads based on the current system load and processing queue length to ensure efficient resource utilization.

[0058] The system sequentially extracts paragraphs P2 through Pn from the paragraph queue Q, encapsulating them as task units and submitting them to the thread pool. Task scheduling utilizes a priority queue mechanism, with earlier paragraphs receiving higher processing priority. Each worker thread retrieves a paragraph processing task from the task queue and executes the same process as the first paragraph processing: semantic analysis, audio generation, image generation, and association establishment.

[0059] To optimize resource usage, the system uses different resource allocation strategies for different types of processing tasks. Text processing and audio generation tasks prioritize CPU resources, while image generation tasks prioritize GPU resources. The system monitors the execution status and progress of each task in real time and pushes progress information to the front-end via WebSocket, allowing users to understand the status of backend processing.

[0060] The AI ​​storybook content of each completed paragraph is uploaded to the cloud storage platform and the corresponding status mark is updated. The system maintains a processing status table that records the processing stage, completion time, and resource link of each paragraph, providing a query basis for the front-end.

[0061] In the embodiment of this step, the first paragraph priority response mechanism adopts a dual-channel mode of "first paragraph instant rendering + background pre-generation"; the front end immediately starts the player component after receiving the AI ​​generation result of the first paragraph of text (including graphic and text layout data), and calculates the subsequent content loading buffer period through the timeline prediction algorithm.

[0062] A dynamic thread pool control system is used to establish a task queue based on the semantic segmentation of story text paragraphs. The thread pool controller monitors in real time: GPU memory usage, logical dependencies between paragraphs, and network fluctuations on user terminals. A priority queue-jumping algorithm is used to handle sudden paragraph rewriting requests.

[0063] For example, when a user creates a 10-page picture book, page 1 plays immediately, and pages 2-4 are loaded when the homepage plays to 70%.

[0064] Step 5: When the first paragraph of the AI ​​story picture book is finished playing, the subsequent paragraphs of the AI ​​story picture book will be played in sequence.

[0065] Specifically, the front-end application monitors the playback status of the first AI storybook segment. When it detects that playback is about to complete (for example, the remaining playback time is less than 3 seconds), it immediately requests the back-end for the next AI storybook segment. Based on the front-end request and the processing status table, the system returns a link to the next processed segment.

[0066] The front-end application uses progressive rendering technology to process received paragraph content. Progressive rendering includes a preloading mechanism and smooth transition effects. The preloading mechanism pre-loads the resources for the next paragraph while the current paragraph is playing, ensuring no waiting when switching between content. The smooth transition effect uses image fade-in and fade-out and audio crossfades to make the transition between paragraphs smooth and natural.

[0067] The front-end application plays the AI ​​storybook's subsequent sections in the order P2, P3, ..., Pn. During playback, the front-end application continuously monitors network status and resource loading, adjusting the preloading strategy accordingly. If poor network conditions are detected, the system reduces the quality of preloaded resources to prioritize playback continuity.

[0068] If the processing of a paragraph is not yet complete, the front-end will display an appropriate loading animation and prompt information, and send an expedited processing request to the back-end. After receiving the expedited processing request, the back-end will increase the priority of the paragraph in the task queue to speed up the processing process.

[0069] After all segments have been played, the system provides story review, collection, and sharing functions to enhance the user's interactive experience. Users can choose to replay the entire story or generate new story content, forming a complete user experience closed loop.

[0070] In a further example of the present invention, an emotional resonance enhancement module function can be added: for example, a biosensor interface is embedded in the player interface: the user's pupil changes are captured by a camera to analyze the focus of attention, and laughter / exclamations are collected by a microphone to evaluate the emotional resonance; then the subsequent paragraphs are dynamically adjusted: the color saturation of the picture (emotion enhancement), the speed of text appearance (cognitive load adaptation), and the intensity of the background music (atmosphere adjustment).

[0071] Further examples of this invention include the addition of cross-modal memory network functionality, such as building a user-specific "story DNA database" that records every interruption, replay, and fast-forward behavior pattern; analyzes content preference shifts across sessions; and establishes a personalized generation parameter matrix. For example, if it is detected that a user frequently replays the "robot character" segment, the weight of the subsequent science fiction elements will be automatically increased.

[0072] The present invention is further described in detail below through a specific application example: This specific application embodiment can solve the three core problems of existing AI storybook generation technology: 1) The problem of long waiting time on the first screen: users need to wait for all content to be generated (3-5 minutes) before they can start reading; 2) The problem of idle and wasted computing resources: serial processing between paragraphs results in CPU / GPU utilization of less than 40%; 3) The problem of inconsistent interactive experience: there is a lack of generation progress visualization and exception handling mechanism.

[0073] like Figure 2 As shown, a method for generating an AI story picture book based on parallel processing in a specific application embodiment of the present invention includes the following steps: S31. The client initiates a generation request and enters S32.

[0074] S32. The server executes and proceeds to step S33.

[0075] S33. Obtain the story generation request input by the user and perform sensitive word detection, and then proceed to step S34.

[0076] Receiving a story generation request from the user: This step involves the system waiting for the user to enter information related to the story they wish to generate. Users may enter descriptions of the story's theme, plot, character settings, specific scenes, and other aspects, or they may specify specific requirements such as the story's style (e.g., adventure, suspense, romance), length (e.g., short, medium, long), and target audience (e.g., children, teenagers, adults). The system must be able to accurately receive and understand user input, whether in text, voice, or other forms of information, and convert it into a data format that the system can process.

[0077] Regarding sensitive word detection, after receiving the user's input, the system needs to detect sensitive words. Sensitive words refer to words or phrases that may contain illegal, bad, immoral or inappropriate information, such as content related to pornography, religious hatred, political sensitivity, etc. Through sensitive word detection technology, the system will compare the text entered by the user with a pre-set sensitive word library to check whether there are matching sensitive words. If sensitive words are found, the system may take different handling methods, such as refusing to generate stories, prompting users to modify the input content, blocking or replacing sensitive words, etc., to ensure that the generated stories comply with ethical and legal standards, as well as the platform's usage regulations.

[0078] Specific as Figure 5 As shown, the picture book initiated by the user through voice or text input is obtained, and then sensitive words are checked by AI (estimated to take 0.5 seconds), and then enter S34; Generally speaking, the purpose of the step of "obtaining user input for story generation requests and performing sensitive word detection" is to accurately obtain user needs before generating stories, and to ensure the legality and appropriateness of the input content, thereby providing a reliable foundation for the subsequent story generation process.

[0079] S34: Call the large model to generate a complete story and proceed to step S35.

[0080] In this step, if Figure 5 As shown, the large model is called to generate the complete story text (the content Qwen2.5-72B is a language model, which can be directly translated as "Tongyi Qianwen 2.5 version - 72 billion parameter model").

[0081] In the embodiment of the present invention, the regular expression "paragraph 1: (.*?)(?=\\s*paragraph[2-9]|$)" can be used to match the first paragraph; and a paragraph queue Q={P1,P2,...,Pn} can be established.

[0082] Among them, a regular expression consists of common characters (such as letters and numbers) and special characters (metacharacters). It defines a rule to check whether a string matches the rule.

[0083] The first paragraph fast extraction mechanism in the embodiment of the present invention adopts regular expression fast matching.

[0084] S35, extract the first paragraph P1 according to the regular expression, and proceed to step S36; S36. Start an independent thread to process PI and proceed to S37.

[0085] In this step, an independent thread T1 is started to process paragraph P1. like Figure 5As shown, an independent thread T1 is started to process paragraph 1, and audio in a specified audio format is generated according to the text content of paragraph 1, and a corresponding image is generated through AI text image according to the text content of paragraph 1; and the image is uploaded to the cloud platform, and a watermark is added.

[0086] S37. Return the P1 content to the client and proceed to step S38.

[0087] In this step, the processing result of the first segment P1 is returned to the front end within 12 seconds, so that the front end can play the result in time and reduce the user's waiting time. In this embodiment of the present invention, the first segment is returned first, and the perceived waiting time is reduced. Figure 5 As shown, in this specific embodiment, the first segment of the time limit is 0.5 seconds (s) for sensitive word checking; the time for generating the picture book content through the large model is about 5.1 seconds (s); the time for generating the first audio and picture is 13.1 seconds (s), and the total time for the first segment is: 18.7 seconds (s).

[0088] S38. Create a thread pool to process P2-Pn, and then go to step S39.

[0089] In this step, a thread pool Pool = {T2, T3, ..., Tn} is created to dynamically allocate subsequent paragraph processing tasks; in the embodiment of the present invention, subsequent paragraphs are multi-threaded and generated in parallel. Multi-threaded parallel processing is used; Figure 5 As shown, paragraphs 2 (8.1 seconds), 3 (10.8 seconds), 4 (13.38 seconds), 5 (17.2 seconds), 6 (20.1 seconds), 7 (23.9 seconds), and 8 (26.9 seconds) are processed respectively. This specific embodiment processes the last paragraph, and the sensitive word check takes 0.5 seconds (s). The time for generating the picture book content through the large model is about 26.9 seconds. The last audio and picture generation time is 11.4 seconds, and the time required to complete all subsequent paragraphs is 38.8 seconds, that is, the time required to unify the sensitive words for 0.5 seconds, the time required to generate the subsequent paragraphs in parallel for the large model for 26.9 seconds, and the time required to generate the mp3 and pictures for 11.4 seconds, a total of 38.8 seconds. Therefore, the subsequent duration is equal to the maximum time required for the large model generation and the maximum time required for the mp3 and pictures.

[0090] S39: Push the processing progress in real time and proceed to S40; S40, client-side progressive rendering.

[0091] As can be seen from the above, the embodiment of the present invention provides an AI storybook generation method based on parallel processing. By combining instant response to the first paragraph with multi-threaded parallel processing, the user waiting time is reduced from minutes to seconds. That is, after the first paragraph is processed, it is returned to the front end. After the first paragraph is played, the subsequent paragraphs are also generated, which greatly reduces the user's waiting time. The present invention also has the following advantages: 1) Greatly reduced latency: first paragraph response time ≤ 12 seconds (measured data); 2) Improved resource utilization: CPU usage increased by 40%; 3) Optimized user experience: user reading progress is synchronized with background processing progress.

[0092] Exemplary devices like Figure 3 As shown, an embodiment of the present invention provides an AI story picture book generation device based on parallel processing, the device comprising: An acquisition module 310 is used to acquire a story generation request; The text generation module 320 is used to call the AI ​​big model to automatically generate the complete story text according to the story generation request; The first paragraph picture book generation module 330 is used to extract the first paragraph text content of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph text content of the complete story text; The first segment playback control and parallel processing module 340 is used to promptly return the AI ​​story picture book generation processing of the first segment of text content to the front-end for playback; and simultaneously create a thread pool to dynamically allocate processing tasks for subsequent segments, and dynamically generate AI story picture books for the text content of the subsequent segments of the complete story text in sequence; The subsequent playback control module 350 is used to play the AI ​​story picture book of subsequent paragraphs in sequence after the playback of the first paragraph of the text content of the AI ​​story picture book is completed.

[0093] Specifically, the acquisition module includes a user interface component and a request processing component. The user interface component provides a graphical interface, allowing users to submit story generation requests through a text input box or voice input. The request processing component receives user input and performs sensitive word detection on the input content. The sensitive word detection component has a built-in multilingual sensitive word library, supports fuzzy matching and contextual analysis, and can effectively identify inappropriate content. Requests that pass the detection are formatted into a standard request format and passed to the text generation module.

[0094] The text generation module includes a model selection component, a text generation component, and a text processing component. The model selection component selects the most suitable large language model from a pool of preconfigured AI models based on the request characteristics and system load. The text generation component is responsible for calling the selected AI large model API, passing the request parameters, and receiving the generated results. The text processing component performs post-processing on the generated complete story text, including format normalization, paragraph division, and quality assessment. This module uses regular expressions to segment the text into multiple paragraphs and establishes a paragraph queue Q = {P1, P2, ..., Pn} to provide a data foundation for subsequent processing.

[0095] The first paragraph picture book generation module includes a thread management component, an audio generation component, and an image generation component. The thread management component is responsible for creating and managing an independent thread T1 and allocating computing resources for the first paragraph processing. The audio generation component performs semantic analysis on the first paragraph text content P1 and uses the text-to-speech engine to generate natural and fluent audio content. The image generation component extracts visual keywords from the text, constructs image generation prompts, and uses the AI ​​image generation model to create images that match the text. This module also includes an association establishment component, which is responsible for establishing a correspondence between the generated audio and images and uploading them to the cloud storage platform.

[0096] The first-segment playback control and parallel processing module includes a resource push component, a thread pool management component, and a task scheduling component. The resource push component is responsible for pushing the processed first segment of the AI ​​storybook content to the front-end application. The thread pool management component creates and maintains a dynamic thread pool, adjusting the number of threads based on system load. The task scheduling component retrieves subsequent segment content from the segment queue, encapsulates it into task units, and assigns it to the thread pool for processing. This module also includes a status monitoring component that records the processing status and progress of each segment in real time and pushes status information to the front-end.

[0097] The subsequent playback control module includes a resource request component, a rendering control component, and a playback management component. The resource request component requests the next paragraph of the AI ​​storybook content from the backend based on the frontend's playback progress. The rendering control component implements progressive rendering, including resource preloading and smooth transition effects. The playback management component controls content playback according to the paragraph sequence, monitors network status and resource loading, and dynamically adjusts playback strategies. This module also includes an interactive component that provides features such as story review, collection, and sharing to enhance the user experience.

[0098] Each module communicates through a standardized data interface, ensuring the consistency and reliability of data flow. The entire device adopts a microservices architecture design, and each functional module can be independently deployed and expanded, improving the maintainability and scalability of the system.

[0099] Based on the above embodiment, the present invention also provides an intelligent terminal, whose principle block diagram can be shown as follows: Figure 4As shown. The intelligent terminal includes a processor, a memory, a network interface, a display screen, and a database connected via a system bus. The processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, an AI story picture book generation method based on parallel processing is implemented. The database of the intelligent terminal is used to store an AI story picture book generation program based on parallel processing.

[0100] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the solution of the present invention and does not constitute a limitation on the smart terminal to which the solution of the present invention is applied. The specific smart terminal may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0101] In one embodiment, a smart terminal is provided, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations: Obtain a story generation request; based on the story generation request, call the AI ​​big model to automatically generate a complete story text; extract the first paragraph of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph of the complete story text; when the AI ​​story picture book generation processing of the first paragraph of the text content is completed, return it to the front end for playback in a timely manner; and at the same time create a thread pool to dynamically allocate processing tasks for subsequent paragraphs, and dynamically perform AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence; when the AI ​​story picture book playback of the first paragraph of the text content is completed, play the AI ​​story picture books of the subsequent paragraphs in sequence.

[0102] Preferably, the step of obtaining a story generation request includes: obtaining a story generation request input by a user, and performing sensitive word detection on the story generation request input by the user.

[0103] Preferably, the step of calling the AI ​​big model to automatically generate a complete story text according to the story generation request includes: calling the AI ​​big model to automatically generate a complete story text according to the story generation request; and using a regular expression to represent the generated complete story text: "Paragraph 1: (.*?)(?=\\s*paragraph[2-9]|$)" to match the first paragraph; and establishing a paragraph queue Q={P1,P2,...,Pn} for paragraphs 1 to paragraph n respectively.

[0104] Preferably, the steps of extracting the first paragraph of the text content of the complete story text, starting an independent thread, and performing AI story picture book generation processing on the first paragraph of the text content of the complete story text include: extracting the first paragraph of the text content of the complete story text; and controlling the start of an independent thread T1 to perform AI story picture book generation processing on the first paragraph of the text content P1 of the complete story text; when performing AI story picture book generation processing on the first paragraph of the text content P1, first generate the first paragraph of audio in a specified audio format according to the first paragraph of the text content P1; and perform AI text picture processing on the first paragraph of the text content P1 to generate the first paragraph of the text picture, and establish a correspondence between the first paragraph of the audio and the first paragraph of the text picture.

[0105] Preferably, the step of performing AI text image processing based on the first paragraph text content P1 to generate the first paragraph text image, and establishing a correspondence between the first paragraph audio and the first paragraph text image also includes: uploading the first paragraph audio and the first paragraph text image with the established correspondence to the front-end cloud storage platform, and adding a corresponding watermark to the first paragraph text image.

[0106] Preferably, the steps of: when the AI ​​story picture book generation processing of the first paragraph of text content is completed, returning it to the front end for playback in time; and simultaneously creating a thread pool to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence include: when the AI ​​story picture book generation processing of the first paragraph of text content is completed, obtaining the AI ​​story picture book corresponding to the first paragraph of text content, and promptly sending it to the front end for real-time playback; while the AI ​​story picture book corresponding to the first paragraph of text content is playing, creating a thread pool Pool={T2, T3, ..., Tn} in parallel in the background to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence, processing and generating audio and text images of corresponding paragraphs in sequence, and establishing a corresponding relationship between the corresponding audio and text images.

[0107] Preferably, the step of playing the AI ​​story picture book of the subsequent paragraphs in sequence after the playback of the AI ​​story picture book of the first paragraph of text content is completed includes: when the playback of the AI ​​story picture book of the first paragraph of text content is completed, according to the processing progress pushed in real time, the AI ​​story picture book of the subsequent paragraphs that have been processed are obtained from the background in sequence, and the AI ​​story picture book of the subsequent paragraphs are played in sequence after progressive rendering, as described above.

[0108] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

Claims

1. A method for generating an AI story picture book based on parallel processing, characterized in that: include: Get story generation request; Based on the story generation request, the AI ​​big model is called to automatically generate the complete story text; Extract the first paragraph of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph of the complete story text; When the AI ​​story picture book generation processing of the first paragraph of text content is completed, it is promptly returned to the front end for playback; and at the same time, a thread pool is created to dynamically allocate the processing tasks of subsequent paragraphs, and the AI ​​story picture book generation processing is dynamically performed on the text content of the subsequent paragraphs of the complete story text in sequence; When the first paragraph of the AI ​​story picture book is finished playing, the subsequent paragraphs of the AI ​​story picture book will be played in sequence.

2. The AI ​​story picture book generation method based on parallel processing according to claim 1 is characterized in that: The step of obtaining a story generation request includes: Obtain a story generation request input by a user, and perform sensitive word detection on the story generation request input by the user.

3. The AI ​​story picture book generation method based on parallel processing according to claim 1 is characterized in that: The steps of calling the AI ​​big model to automatically generate the complete story text according to the story generation request include: Based on the story generation request, the AI ​​big model is called to automatically generate the complete story text; The generated complete story text is represented by a regular expression: paragraph 1, paragraph 2 to paragraph n, where paragraph 1 matches the first paragraph; And establish paragraph queues Q={P1,P2,...,Pn} for paragraphs 1 to n respectively.

4. The AI ​​story picture book generation method based on parallel processing according to claim 1 is characterized in that: The steps of extracting the first paragraph of the complete story text, starting an independent thread, and performing AI story picture book generation processing on the first paragraph of the complete story text include: Extracting the first paragraph of the complete story text; And control to start an independent thread T1 to perform AI story picture book generation processing on the first paragraph text content P1 of the complete story text; When performing AI story picture book generation processing on the first paragraph text content P1, first generate the first audio of the specified audio format according to the first paragraph text content P1; And AI text-generated graph processing is performed based on the first paragraph text content P1 to generate the first paragraph text-generated graph, and a corresponding relationship is established between the first paragraph audio and the first paragraph text-generated graph.

5. The AI ​​story picture book generation method based on parallel processing according to claim 4 is characterized in that: After the step of performing AI text-generating graph processing based on the first paragraph text content P1 to generate the first paragraph text-generating graph and establishing a corresponding relationship between the first paragraph audio and the first paragraph text-generating graph, the following steps are further included: The first audio segment and the first text image with established corresponding relationships are uploaded to the front-end cloud storage platform, and a corresponding watermark is added to the first text image.

6. The AI ​​story picture book generation method based on parallel processing according to claim 1 is characterized in that: The steps of promptly returning the AI ​​story picture book generation processing of the first paragraph of text content to the front end for playback when the processing is completed; and simultaneously creating a thread pool to dynamically allocate processing tasks for subsequent paragraphs, and dynamically performing AI story picture book generation processing on the text content of the subsequent paragraphs of the complete story text in sequence include: When the AI ​​story picture book generation process of the first paragraph of text content is completed, the AI ​​story picture book corresponding to the first paragraph of text content is obtained and sent to the front end in time for real-time playback; While the AI ​​story picture book corresponding to the first paragraph of text content is being played, a thread pool is created in parallel in the background to dynamically allocate processing tasks for subsequent paragraphs, and the text content of the paragraphs following the complete story text is dynamically generated by the AI ​​story picture book. The audio and text images of the corresponding paragraphs are generated in turn, and a corresponding relationship is established between the corresponding audio and text images.

7. The AI ​​story picture book generation method based on parallel processing according to claim 1 is characterized in that: When the AI ​​story picture book of the first paragraph of text content is played, the steps of playing the AI ​​story picture book of subsequent paragraphs in sequence include: When the AI ​​story picture book with the first paragraph of text content is finished playing, the AI ​​story picture book of the subsequent paragraphs that have been processed are obtained from the background in sequence according to the processing progress pushed in real time, and the AI ​​story picture books of the subsequent paragraphs are played in sequence after progressive rendering.

8. An AI story picture book generation device based on parallel processing, characterized in that: The device comprises: The acquisition module is used to obtain story generation requests; The text generation module is used to call the AI ​​model to automatically generate the complete story text based on the story generation request; A first paragraph picture book generation module is used to extract the first paragraph text content of the complete story text, start an independent thread, and perform AI story picture book generation processing on the first paragraph text content of the complete story text; The first paragraph playback control and parallel processing module is used to promptly return the AI ​​story picture book generation processing of the first paragraph of text content to the front-end for playback; and at the same time, create a thread pool to dynamically allocate the processing tasks of subsequent paragraphs, and dynamically generate AI story picture books for the text content of the subsequent paragraphs of the complete story text in sequence; The subsequent playback control module is used to play the AI ​​story picture book of subsequent paragraphs in sequence after the AI ​​story picture book of the first paragraph of text content is played.

9. An intelligent terminal, characterized in that: The device comprises a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, and the one or more programs include being used to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 7.

Images