Method and system for automatically generating novel promotion video based on multi-modal large model

Abstract

The invention provides an automatic generation method and system for a novel promotion video based on a multi-modal large model in the technical field of digital content creation, and the method comprises the following steps: S1, extracting a novel text through analyzing a URL link, a TXT format, a DOCX format or a PDF format document, performing analysis and segmented structured processing on the novel text through an NLP technology to obtain structured data; s2, performing compliance check and optimization on the structured data to obtain compliance data; s3, analyzing and extracting the compliance data based on a thinking chain to generate a shot script; s4, acquiring subtitles and picture cues based on the shot script, synthesizing an audio based on the subtitles, and generating a picture group based on the picture cues; and S5, automatically generating a promotion video of the novel text based on the audio and the picture group. The novel promotion video generation method has the advantages that the novel promotion video generation efficiency is greatly improved, and the novel promotion video generation cost is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of digital content creation, and in particular to a method and system for automatically generating a promotional video for a novel based on a multimodal large model. Background Art

[0002] With the advent of the digital age, novels, as a traditional literary form, are facing unprecedented challenges. With the popularity of social media, video platforms and mobile devices, people's attention is increasingly dispersed. These platforms have attracted a large number of users with their rich multimedia content and high interactivity, making it difficult for traditional text-based novel promotion methods to attract and maintain the attention of potential readers. In addition, the digital age has put forward higher requirements for the immediacy, personalization and visual appeal of content. This transformation requires that the promotion strategy of novels must be innovative to adapt to the modern media environment of dual visual and auditory stimulation.

[0003] As an emerging marketing tool, promotional videos can take advantage of the visual and auditory effects of short video platforms to bring viewers a more intuitive and engaging experience. However, the process of converting novels from text form into promotional videos is complicated and costly. First, it is necessary to conduct in-depth analysis of the key plots, characters and themes of the novel. This process is not only time-consuming and costly, but also usually requires the participation of professional editors. Secondly, in order to enhance the attractiveness of promotional videos, it is often necessary to hire professional voice actors to dub multiple characters, which further increases production costs. Next, the visualization of the novel content, such as through storyboard comics, relies on the creativity and professional skills of designers, which also requires considerable time and resources. In addition, producers also need to conduct compliance reviews of promotional videos in accordance with copyright laws and platform regulations to prevent legal risks and economic losses.

[0004] Therefore, how to provide a method and system for automatically generating novel promotion videos based on a multimodal large model to improve the efficiency of novel promotion video generation and reduce the cost of novel promotion video generation has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for automatically generating novel promotion videos based on a multimodal large model, so as to improve the efficiency of generating novel promotion videos and reduce the cost of generating novel promotion videos.

[0006] In a first aspect, the present invention provides a method for automatically generating a novel promotion video based on a multimodal large model, comprising the following steps:

[0007] Step S1, extracting novel text by parsing URL links, documents in TXT format, DOCX format or PDF format, and performing structural processing on the novel text by parsing and segmenting through NLP technology to obtain structured data;

[0008] Step S2: performing compliance verification and optimization on the structured data to obtain compliant data;

[0009] Step S3: analyzing and extracting the compliance data based on the thought chain to generate a shot script;

[0010] Step S4, obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts;

[0011] Step S5: automatically generate a promotional video of the novel text based on the audio and picture group.

[0012] Furthermore, the step S2 is specifically as follows:

[0013] A sensitive word library including several sensitive words is created, and the sensitive word library is regularly updated through crawler technology. The sensitive words contained in the structured data are first filtered through the sensitive word library, and then the sensitive content contained in the structured data is identified through a large language model, and the identified sensitive content is replaced, thereby completing the compliance verification and optimization of the structured data to obtain compliant data.

[0014] Furthermore, the step S3 is specifically as follows:

[0015] Constructing workflow prompt words based on thought chains, and then analyzing key plots in the compliance data through a large language model to obtain an overview script corresponding to the structure, plot, character development, and theme in the compliance data, extracting key plots including at least turning points, climax points, conflict points, and resolution points from the compliance data based on the overview script, and evaluating the importance of each of the key plots through a large language model to obtain an evaluation result;

[0016] Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

[0017] Furthermore, the step S4 is specifically as follows:

[0018] Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology.

[0019] Furthermore, the step S5 is specifically as follows:

[0020] Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

[0021] In a second aspect, the present invention provides a system for automatically generating a novel promotion video based on a multimodal large model, comprising the following modules:

[0022] A novel text acquisition module is used to extract novel text by parsing URL links, documents in TXT format, DOCX format or PDF format, and to parse and segment the novel text through NLP technology to obtain structured data;

[0023] A compliance checking module, used to perform compliance checking and optimization on the structured data to obtain compliant data;

[0024] A shot script generation module, used to analyze and extract the compliance data based on the thought chain to generate a shot script;

[0025] An audio and picture generation module, used for obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts;

[0026] A promotional video generation module is used to automatically generate a promotional video of the novel text based on the audio and picture group.

[0027] Furthermore, the compliance verification module is specifically used to:

[0028] A sensitive word library including several sensitive words is created, and the sensitive word library is regularly updated through crawler technology. The sensitive words contained in the structured data are first filtered through the sensitive word library, and then the sensitive content contained in the structured data is identified through a large language model, and the identified sensitive content is replaced, thereby completing the compliance verification and optimization of the structured data to obtain compliant data.

[0029] Furthermore, the lens script generation module is specifically used for:

[0030] Constructing workflow prompt words based on thought chains, and then analyzing key plots in the compliance data through a large language model to obtain an overview script corresponding to the structure, plot, character development, and theme in the compliance data, extracting key plots including at least turning points, climax points, conflict points, and resolution points from the compliance data based on the overview script, and evaluating the importance of each of the key plots through a large language model to obtain an evaluation result;

[0031] Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

[0032] Furthermore, the audio picture generation module is specifically used for:

[0033] Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology.

[0034] Furthermore, the promotional video generation module is specifically used for:

[0035] Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

[0036] The advantages of the present invention are:

[0037] Extract novel text by parsing URL links, TXT, DOCX or PDF documents, parse and segment the novel text through NLP technology to obtain structured data, perform compliance check and optimization on the structured data to obtain compliant data, then analyze and extract the compliant data based on the thought chain to generate a shot script, then obtain subtitles and screen prompts based on the shot script, synthesize audio based on the subtitles, generate a picture group based on the screen prompts, and finally automatically generate a promotional video for the novel text based on the audio and picture group; in the process of generating the shot script, the reasoning ability of the large language model is enhanced based on the thought chain, so as to effectively extract the key plot, and then generate the shot script based on the important key plot, without the participation of professional editors; synthesize the audio from the subtitles obtained from the shot script through the ChatTTS technology, without the need to hire professional voice actors to dub multiple roles; through Story Diffus ion technology converts the screen prompt words obtained from the shot script into a group of pictures. This process ensures that the character maintains a consistent identity and clothing features in different pictures without relying on the designer's creativity and professional skills. Combined with sensitive word library filtering and compliance verification and optimization of sensitive content replacement, it avoids the generated promotional video from being regenerated due to compliance issues. The entire process is automatically executed without human participation, which ultimately greatly improves the efficiency of novel promotion video generation and greatly reduces the cost of novel promotion video generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described below in conjunction with embodiments with reference to the accompanying drawings.

[0039] Figure 1 It is a flow chart of a method for automatically generating a promotional video for a novel based on a multimodal large model according to the present invention.

[0040] Figure 2 It is a structural schematic diagram of an automatic generation system of a novel promotion video based on a multimodal large model according to the present invention. DETAILED DESCRIPTION

[0041] The technical solution in the embodiments of the present application has the following overall idea: in the process of generating shot scripts, the reasoning ability of the large language model is enhanced based on the thought chain to effectively extract key plots, and then the shot scripts are generated based on the important key plots without the participation of professional editors; the subtitles obtained from the shot scripts are synthesized into audio through the ChatTTS technology, without the need to hire professional voice actors to dub multiple characters; the screen prompt words obtained from the shot scripts are converted into picture groups through the Story Diffusion technology. This process ensures that the characters maintain consistent identities and clothing features in different pictures without relying on the creativity and professional skills of the designer; combined with the compliance verification and optimization of sensitive word library filtering and sensitive content replacement, the generated promotional video is avoided from being regenerated due to compliance issues. The entire process is automatically executed without human participation, so as to improve the efficiency of novel promotion video generation and reduce the cost of novel promotion video generation.

[0042] Please refer to Figure 1 to Figure 2 As shown, a preferred embodiment of the method for automatically generating a novel promotion video based on a multimodal large model of the present invention comprises the following steps:

[0043] Step S1, extracting novel text by parsing URL links, TXT format, DOCX format or PDF format documents, parsing and segmenting the novel text through NLP technology to obtain structured data; when implemented, it is not limited to the above formats; the present invention is compatible with and supports a variety of different file formats to ensure that novel texts from various sources can be seamlessly integrated; through structured processing, subsequent processing and analysis are facilitated, the generation efficiency of promotional videos is improved, and a solid foundation for generating high-quality promotional videos is laid;

[0044] Step S2: performing compliance verification and optimization on the structured data to obtain compliant data;

[0045] Step S3: analyzing and extracting the compliance data based on the thought chain to generate a shot script;

[0046] Step S4, obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts;

[0047] Step S5: automatically generate a promotional video of the novel text based on the audio and picture group, that is, transform the narrative art of the novel text into a dual experience of vision and hearing, providing an immersive story enjoyment for the audience.

[0048] The step S2 is specifically as follows:

[0049] Create a sensitive word library including several sensitive words, and regularly update the sensitive word library through crawler technology to ensure the timeliness of sensitive words. First, filter the sensitive words contained in the structured data through the sensitive word library, and then identify the sensitive content contained in the structured data through the large language model (LLM), replace the identified sensitive content, and then complete the compliance verification and optimization of the structured data to obtain compliant data. The large language model has excellent semantic analysis and understanding capabilities. Sensitive content is violent content, adult content, and politically sensitive content.

[0050] Compliance verification is to ensure that promotional videos comply with relevant laws and regulations as well as the regulations and standards for content release on various platforms, to avoid legal risks and economic losses, and to avoid limiting the promotional effect of promotional videos due to copyright and compliance issues.

[0051] The step S3 is specifically as follows:

[0052] Based on the chain of thought (COT), a workflow prompt word is constructed, and then the key plots in the compliant data are analyzed through a large language model to obtain an overview script corresponding to the structure, plot, character development and theme in the compliant data. Based on the overview script, key plots including at least turning points, climax points, conflict points and resolution points are extracted from the compliant data, and the importance of each key plot is evaluated through a large language model to obtain an evaluation result; the importance evaluation can be based on the contribution of the key plot to the overall story advancement and its potential in attracting audience interest, including the emotional depth of the key plot, the position in the story development, and the impact on the relationship between characters and the story climax;

[0053] Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

[0054] Given that novel texts are usually long, in order to attract audiences within a limited time, it is necessary to extract coherent and key plots from the novel text, and use the thinking chain to decompose and extract the key plots in the novel text. By breaking down the steps of text analysis and extraction, the large language model gradually extracts the content, ensuring the coherence of the story line and enhancing the attractiveness of the content to the audience.

[0055] The step S4 is specifically as follows:

[0056] Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology. The screen prompts are used to extract storyboards.

[0057] ChatTTS has been trained on more than 100,000 hours of real audio data in Chinese and English, showing excellent naturalness and voice quality. It can cleverly process English or numbers in the text, avoid unnatural pronunciation, ensure the fluency of the voice, support multi-speaker mode, and can accurately control details such as laughter, pauses and modal particles, adding vividness to the synthesized speech. In addition, ChatTTS has extremely rich voice and intonation selection functions, allowing users to select voices that match their personality and emotional atmosphere for each character in the novel, achieving a high degree of customization of the voice. This customized voice output not only enhances the expressiveness of the character, but also effectively enhances the interest of the audience by providing a realistic auditory experience. Therefore, ChatTTS significantly improves the attractiveness and dissemination effect of promotional videos, bringing significant audio-visual advantages to the market promotion of novel texts.

[0058] Story Diffusion uses a consistent self-attention mechanism and uses reference images to guide self-attention calculations, effectively improving the consistency of generated images. This innovative method has achieved remarkable results in maintaining character identity and clothing consistency, providing strong support for the production of promotional videos and ensuring the coherence and visual appeal of generated content.

[0059] The step S5 is specifically as follows:

[0060] Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

[0061] Sizing is done to ensure that the promotional video maintains a good viewing experience on different playback platforms and devices; interpolation is done to smoothly connect the various images, increase visual fluidity, and make the entire promotional video look more natural and attractive; mixing is done to balance multiple audio tracks to prevent any sound from overwhelming other sounds and affecting the overall auditory effect, such as background music and character dialogue.

[0062] A preferred embodiment of the invention is a system for automatically generating a multi-modal large-scale novel promotion video, comprising the following modules:

[0063] The novel text acquisition module is used to extract the novel text by parsing URL links, TXT format, DOCX format or PDF format documents, and parse and segment the novel text through NLP technology to obtain structured data; when implemented, it is not limited to the above formats; the present invention is compatible with and supports a variety of different file formats to ensure that novel texts from various sources can be seamlessly integrated; through structural processing, subsequent processing and analysis are facilitated, the generation efficiency of promotional videos is improved, and a solid foundation for generating high-quality promotional videos is laid;

[0064] A compliance checking module, used to perform compliance checking and optimization on the structured data to obtain compliant data;

[0065] A shot script generation module, used to analyze and extract the compliance data based on the thought chain to generate a shot script;

[0066] An audio and picture generation module, used for obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts;

[0067] The promotional video generation module is used to automatically generate a promotional video of the novel text based on the audio and picture group, that is, to transform the narrative art of the novel text into a dual experience of vision and hearing, providing an immersive story enjoyment for the audience.

[0068] The compliance verification module is specifically used for:

[0069] Create a sensitive word library including several sensitive words, and regularly update the sensitive word library through crawler technology to ensure the timeliness of sensitive words. First, filter the sensitive words contained in the structured data through the sensitive word library, and then identify the sensitive content contained in the structured data through the large language model (LLM), replace the identified sensitive content, and then complete the compliance verification and optimization of the structured data to obtain compliant data. The large language model has excellent semantic analysis and understanding capabilities. Sensitive content is violent content, adult content, and politically sensitive content.

[0070] Compliance verification is to ensure that promotional videos comply with relevant laws and regulations as well as the regulations and standards for content release on various platforms, to avoid legal risks and economic losses, and to avoid limiting the promotional effect of promotional videos due to copyright and compliance issues.

[0071] The lens script generation module is specifically used for:

[0072] Based on the chain of thought (COT), a workflow prompt word is constructed, and then the key plots in the compliant data are analyzed through a large language model to obtain an overview script corresponding to the structure, plot, character development and theme in the compliant data. Based on the overview script, key plots including at least turning points, climax points, conflict points and resolution points are extracted from the compliant data, and the importance of each key plot is evaluated through a large language model to obtain an evaluation result; the importance evaluation can be based on the contribution of the key plot to the overall story advancement and its potential in attracting audience interest, including the emotional depth of the key plot, the position in the story development, and the impact on the relationship between characters and the story climax;

[0073] Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

[0074] Given that novel texts are usually long, in order to attract audiences within a limited time, it is necessary to extract coherent and key plots from the novel text, and use the thinking chain to decompose and extract the key plots in the novel text. By breaking down the steps of text analysis and extraction, the large language model gradually extracts the content, ensuring the coherence of the story line and enhancing the attractiveness of the content to the audience.

[0075] The audio picture generation module is specifically used for:

[0076] Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology. The screen prompts are used to extract storyboards.

[0077] ChatTTS has been trained on more than 100,000 hours of real audio data in Chinese and English, showing excellent naturalness and voice quality. It can cleverly process English or numbers in the text, avoid unnatural pronunciation, ensure the fluency of the voice, support multi-speaker mode, and can accurately control details such as laughter, pauses and modal particles, adding vividness to the synthesized speech. In addition, ChatTTS has extremely rich voice and intonation selection functions, allowing users to select voices that match their personality and emotional atmosphere for each character in the novel, achieving a high degree of customization of the voice. This customized voice output not only enhances the expressiveness of the character, but also effectively enhances the interest of the audience by providing a realistic auditory experience. Therefore, ChatTTS significantly improves the attractiveness and dissemination effect of promotional videos, bringing significant audio-visual advantages to the market promotion of novel texts.

[0078] Story Diffusion uses a consistent self-attention mechanism and uses reference images to guide self-attention calculations, effectively improving the consistency of generated images. This innovative method has achieved remarkable results in maintaining character identity and clothing consistency, providing strong support for the production of promotional videos and ensuring the coherence and visual appeal of generated content.

[0079] The promotion video generation module is specifically used for:

[0080] Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

[0081] Sizing is done to ensure that the promotional video maintains a good viewing experience on different playback platforms and devices; interpolation is done to smoothly connect the various images, increase visual fluidity, and make the entire promotional video look more natural and attractive; mixing is done to balance multiple audio tracks to prevent any sound from overwhelming other sounds and affecting the overall auditory effect, such as background music and character dialogue.

[0082] In summary, the advantages of the present invention are:

[0083] Extract novel text by parsing URL links, TXT, DOCX or PDF documents, parse and segment the novel text through NLP technology to obtain structured data, perform compliance check and optimization on the structured data to obtain compliant data, then analyze and extract the compliant data based on the thought chain to generate a shot script, then obtain subtitles and screen prompts based on the shot script, synthesize audio based on the subtitles, generate a picture group based on the screen prompts, and finally automatically generate a promotional video for the novel text based on the audio and picture group; in the process of generating the shot script, the reasoning ability of the large language model is enhanced based on the thought chain, so as to effectively extract the key plot, and then generate the shot script based on the important key plot, without the participation of professional editors; synthesize the audio from the subtitles obtained from the shot script through the ChatTTS technology, without the need to hire professional voice actors to dub multiple roles; through Story Diffus ion technology converts the screen prompt words obtained from the shot script into a group of pictures. This process ensures that the character maintains a consistent identity and clothing features in different pictures without relying on the designer's creativity and professional skills. Combined with sensitive word library filtering and compliance verification and optimization of sensitive content replacement, it avoids the generated promotional video from being regenerated due to compliance issues. The entire process is automatically executed without human participation, which ultimately greatly improves the efficiency of novel promotion video generation and greatly reduces the cost of novel promotion video generation.

[0084] Although the specific implementation modes of the present invention are described above, those skilled in the art should understand that the specific implementation modes described are only illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for automatically generating a novel promotion video based on a multimodal large model, characterized by: The steps include: Step S1, extracting novel text by parsing URL links, documents in TXT format, DOCX format or PDF format, and performing structural processing on the novel text by parsing and segmenting through NLP technology to obtain structured data; Step S2: performing compliance verification and optimization on the structured data to obtain compliant data; Step S3: analyzing and extracting the compliance data based on the thought chain to generate a shot script; Step S4, obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts; Step S5: automatically generate a promotional video of the novel text based on the audio and picture group.

2. The method for automatically generating a promotional video for a novel based on a multimodal large model as claimed in claim 1, characterized in that: The step S2 is specifically as follows: A sensitive word library including several sensitive words is created, and the sensitive word library is regularly updated through crawler technology. The sensitive words contained in the structured data are first filtered through the sensitive word library, and then the sensitive content contained in the structured data is identified through a large language model, and the identified sensitive content is replaced, thereby completing the compliance verification and optimization of the structured data to obtain compliant data.

3. The method for automatically generating a promotional video for a novel based on a multimodal large model as claimed in claim 1, characterized in that: The step S3 is specifically as follows: Constructing workflow prompt words based on thought chains, and then analyzing key plots in the compliance data through a large language model to obtain an overview script corresponding to the structure, plot, character development, and theme in the compliance data, extracting key plots including at least turning points, climax points, conflict points, and resolution points from the compliance data based on the overview script, and evaluating the importance of each of the key plots through a large language model to obtain an evaluation result; Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

4. The method for automatically generating a promotional video for a novel based on a multimodal large model as claimed in claim 1, characterized in that: The step S4 is specifically as follows: Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology.

5. The method for automatically generating a promotional video for a novel based on a multimodal large model as claimed in claim 1, characterized in that: The step S5 is specifically as follows: Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

6. A system for automatically generating promotional videos for novels based on multimodal large models, characterized by: Includes the following modules: A novel text acquisition module is used to extract novel text by parsing URL links, documents in TXT format, DOCX format or PDF format, and to parse and segment the novel text through NLP technology to obtain structured data; A compliance checking module, used to perform compliance checking and optimization on the structured data to obtain compliant data; A shot script generation module, used to analyze and extract the compliance data based on the thought chain to generate a shot script; An audio and picture generation module, used for obtaining subtitles and screen prompts based on the shot script, synthesizing audio based on the subtitles, and generating a picture group based on the screen prompts; A promotional video generation module is used to automatically generate a promotional video of the novel text based on the audio and picture group.

7. The system for automatically generating promotional videos for novels based on multimodal large models as claimed in claim 6, characterized in that: The compliance verification module is specifically used for: A sensitive word library including several sensitive words is created, and the sensitive word library is regularly updated through crawler technology. The sensitive words contained in the structured data are first filtered through the sensitive word library, and then the sensitive content contained in the structured data is identified through a large language model, and the identified sensitive content is replaced, thereby completing the compliance verification and optimization of the structured data to obtain compliant data.

8. The system for automatically generating promotional videos of novels based on multimodal large models as claimed in claim 6, characterized in that: The lens script generation module is specifically used for: Constructing workflow prompt words based on thought chains, and then analyzing key plots in the compliance data through a large language model to obtain an overview script corresponding to the structure, plot, character development, and theme in the compliance data, extracting key plots including at least turning points, climax points, conflict points, and resolution points from the compliance data based on the overview script, and evaluating the importance of each of the key plots through a large language model to obtain an evaluation result; Based on the evaluation results, key plots are screened, each of the screened key plots is optimized through a large language model, and shot scripts are generated based on each of the optimized key plots.

9. The system for automatically generating promotional videos for novels based on multimodal large models as claimed in claim 6, characterized in that: The audio picture generation module is specifically used for: Based on the shot script, subtitles and screen prompts are obtained, the subtitles are synthesized into audio through ChatTTS technology, and the screen prompts are converted into picture groups through Story Diffusion technology.

10. The system for automatically generating promotional videos of novels based on multimodal large models as claimed in claim 6, characterized in that: The promotion video generation module is specifically used for: Each image in the image group is automatically resized and interpolated, and the audio is automatically mixed. Finally, the image group and the audio are synthesized into a promotional video to automatically generate a promotional video for the novel text.

Images