Method and system for AI script generation and element disassembly based on large language model

Abstract

The invention discloses an AI script generation and element disassembly method and system based on a large language model, and relates to the technical field of computers. The invention aims to assist script creation and disassembly through an artificial intelligence technology and improve the movie production efficiency. In the aspect of script generation, a user quickly builds a story backbone by inputting key information such as a story brief summary and a story type, a keynote, a structure and the like in combination with inspiration and suggestions provided by a large language model, and the story backbone is gradually perfected into an integrated story. A user can flexibly revise, and efficient man-machine cooperation is achieved. In the aspect of script dismantling, the system automatically dismantles elements such as figures, scenes, props and clothes in a script by utilizing a large language model and preset cue words to form detailed description, removes illegal input and redundant information, and finally outputs a standardized Excel table file, so that efficient management and utilization of a script group are facilitated. Human creation skills and AI intelligent assistance are fused, a new way of reducing cost and improving efficiency is brought to movie production, and intelligent development of the movie and television industry is promoted.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for AI script generation and element decomposition based on a large language model. Background Art

[0002] The soul of a movie lies in the story of the script. The birth of a good story depends not only on the expression of the creator, but also on the accurate perception of the audience's psychology. The manual creation of the script takes a long time to complete. The emergence of generative AI technology has made it possible to automate script creation and provide reference opinions for creators, but the existing generative AI often lacks the flexibility to interact with users when creating texts, so it is impossible to introduce the creative skills and insights of human creators. One-time generation is also limited by the number of tokens and cannot generate longer texts, so it lacks practical value. In addition, during the filmmaking process, the collaboration of various departments revolves around the elements (characters, scenes, costumes, props) after the script is disassembled. Although the disassembly of script elements is simple to operate, it plays a vital role. Manual disassembly of long scripts is often time-consuming and labor-intensive, and easy to miss. Summary of the invention

[0003] This method provides a method and system for AI script generation and element decomposition, which realizes efficient human-machine collaboration through the script generation method and quickly produces scripts that incorporate the creative skills and inspiration of human creators. The efficiency of the crew is improved through the script element decomposition method, and AI is used to reduce costs and increase efficiency for film production.

[0004] In a first aspect, the present invention provides a method for AI script generation and element decomposition based on a large language model, including a script generation task and a script decomposition task, specifically comprising the following steps:

[0005] 1) Story concept determination: Users briefly outline the core of the story, including the protagonist, conflict and challenge, and select the type of story. The big language model provides story inspiration and detail suggestions that match the type based on the user input to enrich the story concept.

[0006] 2) Story tone and script structure planning: Users use adjectives to describe the characteristics of the protagonist and villain and the story goals or settings. The large language model can provide vocabulary suggestions that match the style. The story framework is set according to the metronome of the preset classic script structure (such as three-act play, save the cat metronome, etc.). Users enter the key events of each part according to the script structure and determine the turning point of each key event. The large language model assists in improvement and adjustment.

[0007] 3) Script generation: The large language model fills in and polishes the script paragraph by paragraph based on the story framework and key events set by the user to ensure the coherence of the plot and the distinctness of the characters. The user reviews the generated script and uses the editor to modify or rewrite specified parts. The large language model supports iterative optimization.

[0008] 4) Script input and preprocessing: The user inputs the text content of the script, the system verifies the legality of the input, automatically removes illegal characters and redundant information, and ensures the purity of the script content;

[0009] 5) Element decomposition: The large language model uses preset prompts to decompose the main characters and key scenes from the script, and generates detailed character and scene descriptions. It further decomposes detailed elements such as costumes and props from the script, and forms appropriate detailed descriptions for each character based on the content of the dialogue.

[0010] 6) Standardized output: The large language model removes illegal input and redundant fields, and integrates the disassembled elements and descriptions into a standardized Excel spreadsheet file to facilitate subsequent analysis and management; ensure that the table structure is clear and the field naming is standardized to facilitate user reference and sharing.

[0011] In step 1), the type of the selected story is selected from a preset type library, and a detailed description of the type of the story is preset in a retrieval-augmented generation (RAG) knowledge base.

[0012] In step 2), the classic script structure includes a plurality of preset script nodes, and the user inputs key events at each node to construct a story framework.

[0013] In step 4), the text content of the script entered by the user must comply with the preset format requirements. If the input is empty, it is considered illegal input. The user enters the script through the system input interface.

[0014] In step 5), the preset prompt words are used to guide the AI ​​to accurately extract character and scene information in the script from the perspective of a professional producer, while excluding characters and scenes that are not specifically described in the dialogue.

[0015] In step 6), the standardized Excel spreadsheet file contains a standardized format of information such as script scenes, characters, locations, character costumes, props, etc.; the Excel spreadsheet header is: scene x, characters (multiple characters are separated by ","), locations, character costumes (using [] to indicate the relationship with the characters), props (multiple props are separated by ",")

[0016] In a second aspect, the present invention provides a system for AI script generation and element decomposition based on a large language model, including: a script generation subsystem and a script decomposition subsystem;

[0017] The script generation subsystem includes a script gene and tone determination module, a script structure determination module, a script filling module and a script revision module, which are used for each step of the script generation task;

[0018] The script gene and tone determination module is used to determine the core concept and tone of the script;

[0019] The script structure determination module is used to plan the overall structure of the script, receive key events input by the user, and set the story framework according to the preset classic script structure;

[0020] The script filling module is used to fill in the specific content of the script and generate a complete script based on the story framework and key events set by the user;

[0021] The script revision module is used to revise and improve the script, accept the parts modified or rewritten by the user, and support iterative optimization;

[0022] The script disassembly subsystem includes a script input module, a script element extraction and description generation module, a script sub-element extraction and description generation module, and a script element table generation module, which are used to execute various steps of the script disassembly task;

[0023] The script input module is used to read the script content input by the user, verify the legality of the input, and automatically remove illegal characters and redundant information;

[0024] The script element extraction and description generation module uses a large language model to extract key elements from the script, such as roles and scenes, according to preset prompts, and generates corresponding descriptions;

[0025] The script sub-element extraction and description generation module is used to further refine and extract more specific elements, such as costumes, props, etc., based on the script element extraction, and generate corresponding descriptions;

[0026] The script element table generation module is used to organize the extracted script elements and their descriptions into a standardized Excel spreadsheet file. The spreadsheet file contains the standardized format of script scenes, characters, scenes, props, costumes and other information, and the table header is standardized, which is convenient for users to view, share, and analyze and manage later;

[0027] The script generation subsystem and the script disassembly subsystem both realize their functions by calling the large language model.

[0028] The advantages of the present invention are:

[0029] For the script generation subtask. First, determine the genes of the story. The user enters "a one-sentence story" and selects the type of story to which it belongs. The user can also use the big model to provide some inspiration. Then determine the tone of the story. The user enters an adjective to describe the protagonist, an adjective to describe the villain, and a clear and original goal or setting. For this part, the user can also choose to be inspired by the big language model. Next, determine the structure of the story. Set it according to the metronome of the classic script. The user enters the key events that occurred in each part. The user can also use the big language model to provide inspiration for it. Determine the turning point of each key event. The user enters the turning point of the key event, and can also use the big language model to provide inspiration for it. The big language model fills and polishes the story skeleton built by the user into a complete story. The user reviews and revise the complete story. If you are not satisfied with each scene of the complete story, you can let the big language model re-fill the specified part.

[0030] For the script decomposition subtask. The user inputs the script text. The large language model uses preset prompt words to decompose the characters and scenes. The large language model uses preset prompt words to analyze the script and form a detailed character description for each character. Detailed scene descriptions are formed one by one according to the script dialogue content. The large language model extracts props from the script for each scene, and forms a suitable clothing description for each character according to the dialogue content. Remove illegal input and redundant fields to form a standardized Excel spreadsheet file;

[0031] This method provides a method and system for AI script generation and element decomposition, which realizes efficient human-machine collaboration through the script generation method and quickly produces scripts that incorporate the creative skills and inspiration of human creators. The efficiency of the crew is improved through the script element decomposition method, and AI is used to reduce costs and increase efficiency for film production. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of a method for AI script generation and element decomposition based on a large language model of the present invention.

[0033] Figure 2 It is a structural diagram of a system for AI script generation and element decomposition based on a large language model according to the present invention. DETAILED DESCRIPTION

[0034] This method provides a method and system for AI script generation and element decomposition, which realizes efficient human-machine collaboration through the script generation method and quickly produces scripts that incorporate the creative skills and inspiration of human creators. The efficiency of the crew is improved through the script element decomposition method, and AI is used to reduce costs and increase efficiency for film production.

[0035] Applying natural language processing and generative model technology to the disassembly of script generation elements makes it possible to automate script creation. The top-down hierarchical design and human-machine collaborative design improve the efficiency of generative AI and human collaboration, overcome the "black box nature" of generative AI, and greatly improve the quality and practical value of script generation. This method uses the inductive ability of the large language model to efficiently disassemble the script elements, greatly improving the efficiency of subsequent management.

[0036] Script generation is accomplished by utilizing the MotionAgent model based on Tongyi Qianwen-7B. This model is a deep learning-based video content creation tool that can create scripts, generate movie stills, generate images / videos, and write background music through the provided tool set. The model provides an interface for users to input script requirements and combines the Tongyi Qianwen-7B large language model to generate high-quality content. In addition, through the prompt word optimization link, the natural language input by the user is converted into a standard prompt to ensure that a script with a standard format and similarity can be generated, which is convenient for the subsequent processing of other functions.

[0037] The Tongyi Qianwen language model is used to extract, analyze and summarize the script, breaking it down into individual elements, avoiding the tedious manual analysis. For the results returned by the model, the text analysis method is used to extract and display the result elements separately, with a clear structure and easy to understand.

[0038] Please refer to Figure 1 to Figure 2 As shown, a preferred embodiment of the method for AI script generation and element decomposition based on a large language model of the present invention includes the following steps:

[0039] For script generation subtasks.

[0040] Step S10: Determine the genes of the story. The user inputs a "one-sentence story" and selects the type of story. The user can also use the big model to provide some inspiration.

[0041] The step S10 is specifically as follows:

[0042] The user inputs a "one-sentence story". It needs to include who the protagonist is, what he is fighting against, and what crisis he is facing. The genre of the story is classified according to the genre film classification of "Save the Cat: A Screenwriter's Guide": "There is a monster in the house, the golden fleece, wishes come true, trapped, life changes, friendship, mystery investigation, fools win, under the system, superhero". The detailed description of each genre is pre-placed in the retrieval-augmented generation (RAG) knowledge base.

[0043] For example, a user enters a brief script outline: "There is a couple who keep missing each other. One of them thinks they are seeing each other for the first time, but the other knows it is a reunion after a long separation."

[0044] The preset prompt is "You are a professional scriptwriter. Please use the "Save the Cat" technique to help me complete the script writing step by step. My one-sentence story is: There is a pair of lovers who miss each other repeatedly. The genre classifications are: There is a monster in the house, the golden fleece, the wish comes true, the trouble, the life change, the friendship, the reasoning investigation, the fool wins, the system, and the superhero. Which one do you think it belongs to?" The output is as follows.

[0045] The Big Language Model provides inspiration: Based on the story you provided - a pair of lovers repeatedly miss each other, one of them thinks they are seeing each other for the first time, but the other knows that it is a long-lost reunion - this story best fits the characteristics of the "Dramatic Change" genre.

[0046] Life-changing films often focus on the personal growth and transformation of their characters, using a series of events that prompt the protagonist to reflect on their life, relationships, or goals, and ultimately achieve inner change. In your story, the missed opportunities and reunions between the two lovers provide a strong emotional arc that can be used to explore their respective inner lives and how they have changed over time.

[0047] Of course, if your story involves more elements, such as the quest for an unknown destiny or the search for past memories, it may also involve elements of other genres, such as "The Golden Fleece", which is a treasure hunt adventure, or "mystery investigation" if the story focuses on the mystery of why two people keep missing each other. But overall, "life changes" is the most appropriate classification.

[0048] Step S20: Determine the tone of the story. The user enters an adjective to describe the protagonist, an adjective to describe the villain, and a clear and original goal or setting. This part of the user can also choose to be inspired by the large language model.

[0049] The preset prompt is "You are a professional scriptwriter. Please use the "Save the Cat" technique to help me complete the script writing step by step. My one-sentence story is: There is a pair of lovers who miss each other repeatedly. Next, I need an adjective to describe the protagonist, an adjective to describe the villain, and a clear and original goal or setting. Please help me provide some inspiration." The output is as follows.

[0050] Adjectives describing the protagonist: Nostalgic: The lover who knows that this is a long-lost reunion may have a strong nostalgia for the past, always trying to regain those lost times and feelings. He / she may keep all the memories related to the other person, and even unconsciously look for signs of meeting the other person in daily life.

[0051] Adjectives describing villains or sources of conflict: Oblivious: The other lover who thinks they are meeting for the first time can be described as "oblivious", which does not only mean that he / she has forgotten the past encounter, but can also symbolize the lack of memory caused by his / her busyness in life or trauma. This oblivion can become an implicit barrier between the two people that needs to be overcome.

[0052] Clear and original goal or setting: The opening night of an art exhibition in a city full of memories: This setting provides a romantic and artistic background, symbolizing new beginnings and exploration of the past. The two protagonists come to the exhibition for different reasons, one to find inspiration, the other to find someone who has always been in their heart but seems to have disappeared.

[0053] Step S30: determine the structure of the story, set it according to the metronome of the classic script, and the user inputs the key events that occur in each part. The user can also use the large language model to provide inspiration.

[0054] The step S30 is specifically as follows:

[0055] According to the "Save the Cat" metronome, the script is divided into fifteen parts, and the corresponding script nodes are:

[0056] 1. Opening Image (1) 2. Theme Stated (5) 3. Set-up (1-10) 4. Catalyst (12) 5. Debate (12-25) 6. Breakin to Two (25) 7. B Story (30) 8. Fun and Games (30-35) 9. Midpoint (55) 10. Bad Guys Close in (55-75) 11. All is Lost (75) 12. Dark Night of the Soul (75-85) 13. Break into Three (85) 14. Finale (85-110) 15. Finale Image (110)

[0057] The user chooses to use the large language model to provide inspiration. The preset prompt is "I should set some key events to complete the 90-minute feature film: According to the "Save the Cat" metronome, the script is divided into fifteen parts, and the corresponding script nodes are:

[0058] 1. Opening Image (1) 2. Theme Stated (5) 3. Set-up (1-10) 4. Catalyst (12) 5. Debate (12-25) 6. Breakin to Two (25) 7. B Story (30) 8. Fun and Games (30-35) 9. Midpoint (55) 10. Bad Guys Close in (55-75) 11. All is Lost (75) 12. Dark Night of the Soul (75-85) 13. Break into Three (85) 14. Finale (85-110) 15. Finale Image (110) What key events should be set for each part? The output is as follows:

[0059] Opening Image

[0060] Scene 1: A nostalgic lover wakes up in the morning and looks at the photo wall beside the bed, which is filled with beautiful memories with the forgotten lover.

[0061] Scenario 2: He / she prepares breakfast and reminisces about the times they cooked together, with the song playing in the background being the one that played on their first date.

[0062] Scene 3: On the way to work, he / she passes by the place where they first met, stops and stares, as if he / she can see his / her past self.

[0063] Scenario 4: Upon arriving at the office, he / she receives an invitation from a colleague to attend an upcoming art exhibition that will feature some evocative artworks.

[0064] Theme Stated

[0065] Scene 5: During lunch, the nostalgic lover chats with an old friend and mentions that he has been thinking about the past recently. He says, "Sometimes I feel that fate brings us together, but memory may separate us again."

[0066] Scene 6: The old friend responded that some people and things are hard to get back once they are missed, but some people eventually find their own happiness through persistence.

[0067] Set-up

[0068] Scene 7: Show the nostalgic lover’s daily life, including work, social activities, etc., to show that although his / her life is fulfilling, there is always something missing.

[0069] Scene 8: Introducing the life of a forgotten lover, who is busy with work and occasionally feels empty, but does not know why.

[0070] Scene 9: Shows the forgotten lover's family environment, perhaps the relationship with parents or siblings, revealing his / her personality traits.

[0071] Scene 10: The nostalgic lovers decide to accept their colleague's invitation and start preparing to attend the art exhibition.

[0072] Triggering Events (Catalyst)

[0073] Scene 11: While sorting out old things, a nostalgic lover finds a gift from his forgotten lover - a notebook with both of their names engraved on it, which records their travel plans.

[0074] Scene 12: This notebook inspires him / her to find the forgotten lover and start searching for information about the other party on the Internet.

[0075] Start a discussion

[0076] Scene 13: The nostalgic lovers meet up with friends and share their findings and ideas. Some of their friends support them, while others express doubts, sparking a heated discussion about whether they should look for the past.

[0077] Scene 14: After returning home, the nostalgic lovers are alone in their thoughts, reviewing the history of their relationship and weighing the pros and cons.

[0078] Scene 15: I write a letter to my forgotten lover, but I am hesitant about whether to send it. Finally I decide to go to the art exhibition to see what the situation is.

[0079] Break into Two

[0080] Scene 16: The nostalgic lovers dress up and prepare to go to an art exhibition. They encounter a series of small challenges on the way (such as traffic jams and unable to find the entrance), symbolizing their inner uncertainty and fear.

[0081] Scene 17: Finally entering the exhibition, I was attracted by the various artworks, but also felt an inexplicable tension.

[0082] B Story

[0083] Scene 18: At the exhibition, the nostalgic lovers met an old friend who was also looking for love. The two exchanged experiences and provided emotional support.

[0084] Scenario 19: Alternatively, a rival who is interested in the forgotten lover is encountered and the rival tries to prevent the nostalgic lover from approaching the forgotten lover.

[0085] Fun and Games

[0086] Scene 20: A nostalgic lover sees many artworks related to their shared experiences at the exhibition, and each viewing brings back those wonderful moments.

[0087] Scene 21: We passed each other twice, once in front of the exhibits and the other in the rest area. Each approach brought new hope and disappointment.

[0088] Scene 22: The nostalgic lover tries to contact the forgotten lover through social media or other means, but all fails, which adds to the interest and tension of the plot.

[0089] Midpoint

[0090] Scene 23: The nostalgic lover unexpectedly heard the voice of the forgotten lover in a corner of the exhibition, turned around, and the two finally met face to face.

[0091] Scene 24: This is an emotionally charged dialogue scene where the nostalgic lover tries to awaken the other's memory, but the oblivious lover appears confused and overwhelmed.

[0092] Bad Guys Close in

[0093] Scene 25: The forgotten lover has to leave the exhibition early due to an urgent work assignment, leaving the nostalgic lover alone to face the loss.

[0094] Scenario 26: If there is a competitor, he / she may use this moment to express his / her feelings to the forgotten lover, trying to ruin the chances between the two.

[0095] Scene 27: The nostalgic lover tries to catch up with the forgotten lover but loses track in the crowd and feels that all his efforts are in vain.

[0096] All is Lost

[0097] Scene 28: The nostalgic lover returns home, feeling down and that everything is over. He / she puts the notebook and other memories back to where they belong, seemingly ready to give up.

[0098] Scene 29: A man calls an old friend to talk. The friend comforts him / her but also reminds him / her that sometimes things are not as hopeless as they seem.

[0099] Dark Night of the Soul

[0100] Scene 30: The nostalgic lover sits alone in a dark room, surrounded by extinguished candles, symbolizing the despair and self-doubt deep within.

[0101] Scene 31: A person writes a long letter to a forgotten lover, expressing all his / her emotions and regrets, as his / her last attempt to contact the other person.

[0102] Scene 32: Considering whether to send the letter, you finally decide to make one last effort anyway.

[0103] Break into Three

[0104] Scene 33: The nostalgic lover decides to take action and make one last effort no matter what the outcome. He / she chooses to wait for the forgotten lover at a place they used to go to.

[0105] Scenario 34: Make a video telling the story of the two people's past, attach the letter, and post it on social media, hoping to attract the attention of the forgotten lover.

[0106] Finale

[0107] Scene 35: After watching the video and the letter, the forgotten lover begins to recall the past and gradually realizes the importance of this relationship.

[0108] Scene 36: The forgotten lover decides to go find the nostalgic lover, and the two meet again in a familiar place, and this time they do not miss each other.

[0109] Scene 37: The two embrace each other affectionately. Everything around them seems to be still, with only the sounds of their heartbeats and breathing.

[0110] Scene 38: They walk out of that place together, hand in hand, towards an unknown but hopeful future.

[0111] Finale Image

[0112] Scene 39: The story ends with a new everyday scene showing the two of them having a happy time together. They walk down the street, the sun shines on them, and music plays softly in the background, echoing the opening scene, showing that time has passed, but they have finally found each other.

[0113] Scene 40: The last shot could be a long shot, with the two standing at a high point in the city, overlooking the entire city, symbolizing that they are ready to face any challenges in the future.

[0114] Step S40: Determine the turning point of each key event. The user inputs the turning point of the key event, and a large language model can also be used to provide inspiration for it.

[0115] The preset prompt is "I need to determine the turning points of each key event so that the story transitions smoothly."

[0116] To ensure that the story transitions smoothly and there are clear turning points between each key event, define a clear "turning point" or "trigger mechanism" for each metronome section that will drive the plot forward while maintaining the emotional coherence of the characters and the interest of the audience. Here are detailed suggestions for the key events and their turning points for your 90-minute feature film:

[0117] Opening Image

[0118] Turning point: The nostalgic lover wakes up in the morning and sees a special photo on the photo wall that reminds him / her of the moment he / she first met the forgotten lover. This memory becomes the driving force of his / her behavior that day.

[0119] Theme Stated

[0120] Turning point: During a conversation between the nostalgic lover and an old friend during lunch, the old friend mentioned: "Sometimes you feel that fate brings you together, but memory may separate you again." This sentence deeply touched the protagonist and became a turning point in his / her thinking and actions.

[0121] Set-up

[0122] Turning point: The nostalgic lover receives an invitation from a colleague to attend an art exhibition that will feature some evocative artworks. This invitation becomes an opportunity for him / her to rediscover the forgotten lover.

[0123] Triggering Events (Catalyst)

[0124] Turning point: When sorting out old things, the nostalgic lover found a notebook given to him by his forgotten lover, which recorded their travel plans. This notebook inspired his / her determination to find the other party and became the key driving force to enter the next stage.

[0125] Start a discussion

[0126] Turning point: During a discussion at a gathering with friends, the nostalgic lover decided to write a letter to the forgotten lover, but hesitated whether to send it. In the end, a friend's words encouraged him / her to go to the art exhibition to see the situation first, which determined his / her next move.

[0127] Break into Two

[0128] Turning Point: On the way to the art exhibition, the nostalgic lover encounters a series of small challenges, symbolizing inner uncertainty and fear. However, each challenge is overcome and he / she successfully reaches the exhibition, and this series of challenges becomes a turning point in his / her determination.

[0129] B Story

[0130] Turning point: During the exhibition, the nostalgic lover meets an old friend who is also looking for love, or a competitor who is interested in the forgotten lover. This encounter not only provides emotional support or conflict, but also motivates him / her to find the forgotten lover more determinedly.

[0131] Fun and Games

[0132] Turning point: The nostalgic lover passed by the forgotten lover many times in the exhibition, and each approach was a new hope and disappointment. When they passed by each other for the last time, the nostalgic lover realized that he had to take a more proactive approach, such as contacting each other through social media or other means.

[0133] Midpoint

[0134] Turning point: The nostalgic lover accidentally hears the voice of the forgotten lover in a corner of the exhibition, and the two finally come face to face. This is an emotional climax, with the nostalgic lover trying to awaken the other's memory, but the forgotten lover appears confused and overwhelmed. This scene becomes the turning point of the entire story, marking the plot's shift from "searching" to "revealing."

[0135] Bad Guys Close in

[0136] Turning point: The forgotten lover has to leave the exhibition early because of an urgent work assignment, leaving the nostalgic lover alone to face the loss. If there is a competitor, he / she may use this moment to express his / her feelings to the forgotten lover, trying to destroy the chance between the two. This series of events makes the nostalgic lover feel that all the efforts have been in vain, but it also inspires him / her to be more determined.

[0137] All is Lost

[0138] Turning point: The nostalgic lover returns home, feeling down and thinking that everything is over. However, while sorting through his / her belongings, he / she finds an unsent letter, which becomes a symbol of his / her last effort and motivates him / her to keep going.

[0139] Dark Night of the Soul

[0140] Turning point: The nostalgic lover sits alone in a dark room surrounded by extinguished candles, symbolizing the despair and self-doubt deep inside. But after writing a long letter, he / she decides to make one last effort anyway, and this decision becomes a turning point for him / her to get out of the darkness.

[0141] Break into Three

[0142] Turning point: The nostalgic lover decides to take action and make one last effort, no matter what the outcome. He / she chooses to wait for the forgotten lover at a place they used to go to and makes a video telling the story of their past. This series of actions indicates that he / she is ready for the final outcome.

[0143] Finale

[0144] Turning point: After watching the video and the letter, the forgotten lover begins to recall bits and pieces of the past and gradually realizes the importance of this relationship. This memory becomes a turning point for him / her to decide to find the nostalgic lover, and the two meet again in a familiar place, and this time they do not miss each other.

[0145] Finale Image

[0146] Turning point: The story ends with a new everyday scene, showing the two people's happy time together. The last shot can be a long shot, with the two standing at a high point in the city, overlooking the entire city, symbolizing that they are ready to face any challenges in the future. This scene not only echoes the beginning, but also a beautiful prospect for the future life of the two.

[0147] Step S50: The large language model fills in and polishes the story skeleton built by the user into a complete story.

[0148] The preset prompt is "Please fill in and polish the story part by part based on the story outline you just built. Complete the first part [Opening Scene] first."

[0149] Opening Image

[0150] Scene 1: Memories of the morning

[0151] The camera moves slowly forward, and the sun shines through the gap in the curtains into a cozy bedroom. The nostalgic lover (we call him Lin Yu) slowly wakes up in the morning light, with a hint of tenderness and nostalgia in his eyes. The photo wall at the head of the bed is covered with various photos, each of which records the beautiful moments of the past.

[0152] Close-up: In one photo, Lin Yu and his forgotten lover (we call her Xiaoxuan) are standing on the beach, the setting sun shining on their faces, and both of them are smiling brightly and happily.

[0153] Lin Yu gently stroked the photo, as if he could feel the temperature at that time. He took a deep breath, as if trying to capture those memories that had gone away.

[0154] Scene 2: Breakfast time

[0155] Lin Yu got up and walked into the kitchen to prepare breakfast. He skillfully fried eggs and cut fruits. The background music was an old song that played on their first date - "Those Flowers". As the melody sounded, Lin Yu's thoughts involuntarily drifted back to the past.

[0156] Flashback: Lin Yu and Xiao Xuan were sitting in a small cafe, with drizzle falling outside the window. They were sharing the same dessert and laughing.

[0157] Reality: Lin Yu smiled, holding a piece of bread in his hand, as if he could see himself back then. He put the breakfast on the table, but suddenly realized that he was the only one enjoying it now.

[0158] Scene 3: Staring on the street

[0159] After breakfast, Lin Yu put on his coat and walked out of the house. He walked on the familiar street and passed the place where they first met - an old bookstore. On the stone steps in front of the bookstore, Xiaoxuan used to sit and read.

[0160] Close-up: Lin Yu stopped in front of the bookstore and stared at the glass door, as if he could see Xiaoxuan sitting inside. His eyes were filled with nostalgia and a hint of sadness.

[0161] Narration (optional): Lin Yu's voice sounded softly, "This is the starting point of our story."

[0162] Scene 4: Exhibition invitation

[0163] Lin Yu continued walking to the subway station, ready to go to work. While waiting for the train, he received a text message from a colleague about an upcoming art exhibition.

[0164] Text message: "Hey, Lin Yu! I heard there's an interesting art exhibition. Do you want to go and have a look? The name of the exhibition is 'Traces of Time'. Maybe we can find some inspiration!"

[0165] Lin Yu looked at the screen of his mobile phone and swiped his fingers gently to browse the information of the exhibition. The theme of the exhibition was "Memory and Rebirth", and the works on display were all artists' review and reflection on their past lives.

[0166] Lin Yu's eyes gradually became firm, as if he was attracted by the exhibition, and an irrepressible curiosity and expectation surged in his heart.

[0167] Ending shot:

[0168] The camera zooms out, and Lin Yu stands on the subway platform. The crowd around him is in a hurry, but his eyes stay somewhere in the distance, as if he is looking for something. At this time, the background music gradually fades, leaving a quiet place.

[0169] Step S60: The user reviews and revises the story. If the user is not satisfied with each scene in the story, the user can ask the large language model to complete the filling of the designated part again.

[0170] The step S60 is specifically as follows:

[0171] Users can view the completed script provided by the big language model and modify it in the editor, or let the big language model complete the modification of the specified paragraph. Users can select a part for the big language model to modify.

[0172] Break down the script into subtasks.

[0173] Step S70: The user inputs the script text content. For example, input the [Opening Scene] script of the previous part:

[0174] Step S80: The large language model uses preset prompt words to break down characters and scenes.

[0175] The step S80 is specifically as follows: the large language model preset prompt words are: "You are a professional producer, you will complete the script disassembly work, the script is given, please extract the characters that appear and the scenes where the events occur. Please note that the characters and scenes mentioned in the dialogue (if there is no related picture) do not need to be extracted. [Script content]"

[0176] The output is:

[0177] figure

[0178] Lin Yu (Nostalgic Lover)

[0179] Xiaoxuan (a forgotten lover who only appears in memories)

[0180] Scenario

[0181] Cozy bedroom

[0182] Description: Sunlight shines into the room through the gap in the curtains, and the photo wall at the head of the bed is covered with photos of Lin Yu and Xiao Xuan. Lin Yu wakes up in bed, gently touches the photos, and recalls the good old days.

[0183] Key elements: photo wall, sunshine, bed, close-up photo (Lin Yu and Xiao Xuan standing on the beach).

[0184] kitchen

[0185] Description: Lin Yu got up and walked into the kitchen, preparing breakfast skillfully. The background music was the old song "Those Flowers" that played on their first date. As the melody sounded, Lin Yu's thoughts drifted back to the past.

[0186] Key elements: fried eggs, cut fruits, bread, and background music "Those Flowers".

[0187] Flashback: Little Cafe

[0188] Description: Lin Yu and Xiao Xuan sat in a small cafe, drizzling outside the window, sharing a dessert and laughing. This was their first date.

[0189] Key elements: small cafe, rainy day, dessert, laughter (Note: this scene is a flashback and does not appear in the real scene).

[0190] street

[0191] Description: After Lin Yu finished his breakfast, he put on his coat, walked out of the house, and walked on the familiar street. He passed by an old bookstore, where Xiaoxuan used to sit on the stone steps at the door, reading.

[0192] Key elements: the street, the old bookstore, the stone steps in front of the bookstore, and the glass door.

[0193] metro station

[0194] Description: Lin Yu came to the subway station and prepared to go to work. While waiting for the train, he received a text message from a colleague about an upcoming art exhibition. Lin Yu browsed the exhibition information and his eyes gradually became firm.

[0195] Key elements: subway platform, crowd, mobile phone, text message content (art exhibition invitation), exhibition theme "Memory and Rebirth".

[0196] Ending shot

[0197] Description: The camera zooms out, and Lin Yu stands on the subway platform. The crowd around him is in a hurry, but his eyes stay somewhere in the distance, as if he is looking for something. The background music gradually fades, leaving a quiet place.

[0198] Key elements: subway platform, crowd, Lin Yu’s gaze, background music.

[0199] Step S90: The large language model analyzes the script using preset prompt words to form detailed character descriptions for each character and detailed scene descriptions based on the script dialogue content.

[0200] Furthermore, the steps S90-S100 are specifically as follows:

[0201] The preset prompt words of the large language model are: "You are a professional producer. You will complete the script disassembly work. The script and related characters are given below. Please summarize the character characteristics and scene characteristics in turn. [Script content][Character list][Scene list]"

[0202] Character traits:

[0203] Lin Yu: Nostalgic, gentle, introspective, and determined. He is full of nostalgia for his past relationships, but is also trying to find new possibilities.

[0204] Xiaoxuan: Gentle and lively, mainly presented through Lin Yu’s memories, symbolizing the good times in the past.

[0205] Scene features:

[0206] A warm bedroom: sunshine, photo wall and bed create an emotional atmosphere of nostalgia and remembrance.

[0207] Kitchen: Breakfast preparation, background music, memories and reality interweaving, highlighting Lin Yu's loneliness.

[0208] Streets: Familiar streets, old bookstores, nostalgia and sadness coexist, reinforcing the longing for the past.

[0209] Subway station: Crowds, text messages, expectations and decisions, setting the stage for the development of the plot.

[0210] Ending shot: subway platform, quiet atmosphere, contemplation and searching, foreshadowing a new journey.

[0211] Step S100: The large language model extracts props from the script for each scene, and forms a suitable clothing description for each character according to the content of the dialogue.

[0212] Scene 1: Cozy bedroom

[0213] Props

[0214] Curtains: Sunlight shines into the room through the gaps in the curtains, which may be gauzy white or beige, creating a soft light effect.

[0215] · Photo wall: The photo wall at the head of the bed is full of photos of Lin Yu and Xiao Xuan. These photos are an important carrier of memories, recording the beautiful moments of their past. One of the close-up photos shows the two standing by the sea, the sunset shining on their faces, and their smiles are bright.

[0216] Bed: Lin Yu wakes up in a bed that is neatly made and may have simple bedding, such as light blue or white sheets and pillows, symbolizing tranquility and comfort.

[0217] Character clothing features

[0218] Lin Yu: Wear loose pajamas or home clothes, the color may be relatively plain, such as light gray or beige, showing the laziness and relaxation of the morning. The hair is a little messy, which looks natural and not deliberate.

[0219] Step S110, remove illegal input and redundant fields to form a standardized Excel spreadsheet file;

[0220] The Excel table headers are: Scene x, Characters (separated by “,” between characters), Location, Character Costumes ([] are used to indicate the relationship with the character), Props (separated by “,” between props)

[0221]

[0222] A preferred embodiment of the system for AI script generation and element decomposition based on a large language model of the present invention includes two subsystems:

[0223] For the script generation subsystem, the following modules are included:

[0224] In the script gene and tone determination module, users input a one-sentence summary of the story and the relevant story type, and refine the detailed description of the story. A large language model can be used to provide some inspiration.

[0225] In the script structure determination module, users input the key events of each part of the script. A large language model can be used to provide some inspiration.

[0226] In the script filling module, the large language model fills in the script for the user part by part, and the user can also fill in the part that he or she is inspired by first.

[0227] Script revision module: Users can revise the script completed by the large language model and specify a certain part for the large language model to rewrite.

[0228] The script disassembly subsystem includes the following modules:

[0229] Script input module. The system reads the script input from the user. If the input is empty, it is illegal and cannot proceed to the next step.

[0230] Script element extraction and description generation module. The large language model extracts elements such as roles and scenes from the script using preset prompts and forms relevant descriptions;

[0231] Script sub-element extraction and description generation module. The large language model further extracts elements such as costumes and props from the script with preset prompts and forms relevant descriptions;

[0232] In the script element table generation module, the system removes illegal input and redundant fields to form a standardized Excel spreadsheet file.

[0233] Furthermore, the script genes and keynotes are specifically determined as follows:

[0234] The user inputs a "one-sentence story". It needs to include who the protagonist is, what he is fighting against, and what crisis he is facing. The genre of the story is classified according to the genre film classification of "Save the Cat: A Screenwriter's Guide": "There is a monster in the house, the golden fleece, wishes come true, trapped, life changes, friendship, mystery investigation, fools win, under the system, superhero". The detailed description of each genre is pre-placed in the retrieval-augmented generation (RAG) knowledge base.

[0235] Furthermore, the script structure determination module is specifically:

[0236] According to the "Save the Cat" metronome, the script is divided into fifteen parts, and the corresponding script nodes are:

[0237] 1. Opening Image (1) 2. Theme Stated (5) 3. Set-up (1-10) 4. Catalyst (12) 5. Debate (12-25) 6. Breakin to Two (25) 7. B Story (30) 8. Fun and Games (30-35) 9. Midpoint (55) 10. Bad Guys Close in (55-75) 11. All is Lost (75) 12. Dark Night of the Soul (75-85) 13. Break into Three (85) 14. Finale (85-110) 15. Finale Image (110)

[0238] Furthermore, the script filling module is specifically:

[0239] The preset prompt words of the large language model are: "You are a professional scriptwriter. Here is the framework of [a story in one sentence]. Please fill it in paragraph by paragraph to form a complete script." Then input the large language model paragraph by paragraph. Get the output.

[0240] Furthermore, the script revision module is specifically:

[0241] Users can view the completed script provided by the big language model and modify it in the editor, or let the big language model complete the modification of the specified paragraph. Users can select a part for the big language model to modify.

[0242] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments 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 AI script generation and element decomposition based on a large language model, characterized in that The steps include: 1) Story concept determination: Users briefly outline the core of the story, including the protagonist, conflict and challenge, and select the type of story. The big language model provides story inspiration and detail suggestions that match the type based on the user input to enrich the story concept. 2) Story tone and script structure planning: Users use adjectives to describe the characteristics of the protagonist and villain and the story goals or settings, and the big language model can provide vocabulary suggestions that match the style; The story framework is set according to the metronome of the preset classic script structure. The user inputs the key events of each part according to the script structure and determines the turning point of each key event. The large language model assists in improvement and adjustment. 3) Script generation: The large language model fills in and polishes the complete script paragraph by paragraph based on the story framework and key events set by the user to ensure the coherence of the plot and the distinct characters; Users review the generated scripts and use the editor to modify or rewrite specific parts. The large language model supports iterative optimization. 4) Script input and preprocessing: The user inputs the text content of the script, the system verifies the legality of the input, automatically removes illegal characters and redundant information, and ensures the purity of the script content; 5) Element decomposition: The large language model uses preset prompts to decompose the main characters and key scenes from the script, and generates detailed character and scene descriptions. It further decomposes detailed elements such as costumes and props from the script, and forms appropriate detailed descriptions for each character based on the content of the dialogue. 6) Standardized output: The large language model removes illegal input and redundant fields, and integrates the disassembled elements and descriptions into a standardized Excel spreadsheet file to facilitate subsequent analysis and management; ensure that the table structure is clear and the field naming is standardized to facilitate user reference and sharing.

2. A method for AI script generation and element decomposition based on a large language model as described in claim 1, characterized in that In step 1), the type of the story is selected from a preset type library, and a detailed description of the type of the story is preset in a search enhancement generation knowledge base.

3. A method for AI script generation and element decomposition based on a large language model as described in claim 1, characterized in that In step 2), the classic script structure includes a plurality of preset script nodes, and the user inputs key events at each node to construct a story framework.

4. A method for AI script generation and element decomposition based on a large language model as described in claim 1, characterized in that In step 4), the text content of the script entered by the user must comply with the preset format requirements. If the input is empty, it is considered illegal input; the user enters the script through the system input interface.

5. The method for AI script generation and element decomposition based on a large language model as described in claim 1, characterized in that In step 5), the preset prompt words are used to guide the AI ​​to accurately extract character and scene information in the script from the perspective of a professional producer, while excluding characters and scenes that are not specifically described in the dialogue.

6. A method for AI script generation and element decomposition based on a large language model as described in claim 1, characterized in that In step 6), the Excel table header of the standardized Excel table file is: scene x, character, location, character costume, prop; multiple characters are separated by ","; character costumes use [] to indicate their affiliation with the character; multiple props are separated by ",".

7. A system for AI script generation and element decomposition based on a large language model, characterized by include: Script generation subsystem and script disassembly subsystem; The script generation subsystem includes a script gene and tone determination module, a script structure determination module, a script filling module and a script revision module, which are used for each step of the script generation task; The script gene and tone determination module is used to determine the core concept and tone of the script; The script structure determination module is used to plan the overall structure of the script, receive key events input by the user, and set the story framework according to the preset classic script structure; The script filling module is used to fill in the specific content of the script and generate a complete script based on the story framework and key events set by the user; The script revision module is used to revise and improve the script, accept the parts modified or rewritten by the user, and support iterative optimization; The script disassembly subsystem includes a script input module, a script element extraction and description generation module, a script sub-element extraction and description generation module, and a script element table generation module, which are used to execute various steps of the script disassembly task; The script input module is used to read the script content input by the user, verify the legality of the input, and automatically remove illegal characters and redundant information; The script element extraction and description generation module uses a large language model to extract key elements from the script, such as roles and scenes, according to preset prompts, and generates corresponding descriptions; The script sub-element extraction and description generation module is used to further refine and extract more specific elements, such as costumes, props, etc., based on the script element extraction, and generate corresponding descriptions; The script element table generation module is used to organize the extracted script elements and their descriptions into a standardized Excel spreadsheet file. The spreadsheet file contains the standardized format of script scenes, characters, scenes, props, costumes and other information, and the table header is standardized, which is convenient for users to view, share, and analyze and manage later; The script generation subsystem and the script disassembly subsystem both realize their functions by calling the large language model.

Images