Method, program, and apparatus for generating context-maintaining images on basis of generative artificial intelligence model

Abstract

The present invention relates to a method, a program, and an apparatus for generating images on the basis of a generative artificial intelligence model, and the method may comprise the steps of: when a script is input on the basis of a user input, dividing the script into units of sentences to extract individual sentences; extracting context information from the individual sentences; forming a context sequence by integrating the context information into the induvial sentences on the basis of the context before and after each sentence, and then editing the script through sentence unit matching using the context sequence; and providing the edited script as a prompt for a generative artificial intelligence model to generate, on the basis of the prompt, images in which the context between sentences is maintained.

Description

Method, program, and device for generating context-preserving images based on a generative artificial intelligence model

[0001] The present invention relates to an image generation method based on a generative artificial intelligence model, and more particularly, to a method for maintaining context between images generated through a generative AI model.

[0002]

[0003] The material described in this section merely provides background information on one embodiment of the present invention and does not constitute prior art.

[0004] Generative AI systems are AI systems that generate results based on a user's specific needs. They learn from data sources and are used to create diverse content, including novels, poetry, images, videos, coding, and art. In these generative AI systems, text / image conversion is a crucial task, transforming text into visually realistic images and can be seen as a key milestone toward achieving human-like AI.

[0005] With the advancement of deep learning technology, which allows data to be learned hierarchically and inferred based on patterns, following the human brain's neural network, converting text into images is becoming one of the most impressive applications in the field of computer vision.

[0006] These generative AI systems could convert scripts into images to automatically generate movie storyboards, transform text stories into webtoon format, or transform product descriptions from plain text into engaging, easy-to-understand images.

[0007] A variety of research has been conducted using generative AI systems to convert text into images. Representative methods include GAN-based, autoregressive, and diffusion-based approaches. Commercially available systems include OpenAI's DALL-E and Google's Party.

[0008] Figure 1 is a diagram briefly explaining the configuration of a generative AI system that generates general text into an image.

[0009] As shown in Fig. 1, when a user inputs a script (Text) through a text input interface, the generative AI model converts the input script (Text) into an image and outputs the converted image through a text output interface. At this time, the script input through the text input interface consists of several sentences (S1, S2, ..., S n ) is composed of.

[0010] Conventional script input is done in text units, making it difficult to distinguish between sentences (S). This presents a problem: when multiple sentences are entered, only one image is input for each sentence. To address this issue, inputting sentences (S) into the existing system creates a problem where the context is lost between S1, S2, ...Sn.

[0011] Figure 2 is a diagram comparing images generated using various generative AI models. The problem of lack of context between sentences is explained through Figure 2 as follows.

[0012] As an example of the input script text in Figure 2, if the text T is T={A man is taking a walk. The man picked up a bag on the road. And then sat down next to the bag}, the user wants a certain context to be maintained between the three sentences. For example, they want to maintain the consistency of the background, the consistency of the characters, etc. However, as can be seen in Figure 2, the current generative AI method does not maintain a certain context between the images generated for each sentence. This problem is the context maintenance problem of generative AI. The context maintenance problem is very important for image-generating AI to be used in actual industrial settings.

[0013] Although the different images generated from each generative AI model accurately represent a given sentence unit, the contextual flow of the original text sentence disappears between the generated images, and this context loss problem makes them unsuitable for use as storyboards or webtoons.

[0014] For example, when generating webtoons by inputting a webtoon scenario, if the resulting images don't follow the same context, each image will have a different concept, making it unusable for service applications. Similarly, when inputting a movie scenario as text, if an image completely out of context is generated for each sentence, it will be unusable in the field. This problem applies equally to image-generating AI as well as video-generating AI.

[0015] While significant progress has been made in converting text scripts and novels into image content using generative AI systems, there are still many challenges for general users to effectively utilize them for the task of converting text into images.

[0016]

[0017] In order to solve the above-mentioned problem, the present invention provides a method for maintaining the context of a generated AI model by integrating context information into each sentence after dividing the script into sentence units and considering the context before and after each sentence when a user inputs a plurality of sentences as a script according to one embodiment of the present invention, and then providing the edited script as a prompt for a generative AI model.

[0018] However, the technical task that this embodiment seeks to achieve is not limited to the technical task described above, and other technical tasks may exist.

[0019]

[0020] As a technical means for achieving the above-described technical task, an embodiment of the present invention provides a method for generating an image based on a generative artificial intelligence model, which is performed by a computing device including at least one processor. The method may include the steps of: when a script based on a user input is input, dividing the script into sentence units and extracting individual sentences; extracting contextual information from the individual sentences; integrating the contextual information into individual sentences based on the pre- and post-context of each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the contextual sequence; and providing the edited script as a prompt for the generative artificial intelligence model to generate an image in which the context between sentences is maintained based on the prompt.

[0021] Alternatively, the step of extracting contextual information from the individual sentences above is to extract contextual information centered on the subject word and verb word through morphological analysis in each sentence.

[0022] Alternatively, based on the context before and after each sentence, the step of integrating the context information into individual sentences to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence is to use an Initial Context First (ICF) method to duplicate the context information of the first sentence separated from the script and add it to all sentences within the paragraph.

[0023] Alternatively, the step of editing the script by integrating the context information into individual sentences based on the pre- and post-context of each sentence to form a context sequence, and then matching sentences using the context sequence, uses a Region Context First (RCF) method to set a sentence threshold value that specifies a range in which context continuity should be maintained, and extracts new context information to reconstruct the context sequence when the sentence threshold value is exceeded.

[0024] Alternatively, the step of integrating the context information into individual sentences based on the pre- and post-context of each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence may include a step of adding the context information by selectively using an Initial Context First (ICF) method and a Region Context First (RCF) method.

[0025] Alternatively, the step of adding the context information by selectively using the Initial Context First (ICF) method and the Region Context First (RCF) method may include the steps of: setting a sentence threshold value specifying a range in which context continuity should be maintained; automatically adding context information of the first sentence to the second sentence when a first sentence in the script is input and a second sentence consecutive to the first sentence is input within a preset time frame if the second sentence is within the sentence threshold value; and extracting new context information to reconstruct a context sequence when the second sentence exceeds the sentence threshold value.

[0026] Meanwhile, according to one embodiment of the present invention, there is provided a computer program stored in a computer-readable storage medium, which, when executed on one or more processors, causes the computer program to perform operations for generating an image based on a generative artificial intelligence model, wherein the operations may include: when a script based on a user input is input, an operation for dividing the script into sentence units and extracting individual sentences; an operation for extracting contextual information from the individual sentences; an operation for integrating the contextual information into individual sentences based on the pre- and post-context of each sentence to form a contextual sequence, and then editing the script through sentence-by-sentence matching using the contextual sequence; and an operation for providing the edited script as a prompt for the generative artificial intelligence model, and generating an image in which the context between sentences is maintained based on the prompt.

[0027]

[0028] According to the problem solving means of the present invention described above, when an image is generated based on a prompt consisting of multiple sentences using a generative AI model, the present invention automatically edits the prompt to include the main context and inputs it into the generative AI model, thereby preserving the context between sentences based on user input and maintaining the context between generated images.

[0029]

[0030] Figure 1 is a diagram briefly explaining the configuration of a generative AI system that generates general text into an image.

[0031] Figure 2 illustrates a contextual problem in a system utilizing existing generative AI.

[0032] FIG. 3 is a block diagram of a computing device according to one embodiment of the present invention.

[0033] FIG. 4 is a drawing illustrating a detailed configuration of a computing device according to one embodiment of the present invention.

[0034] Figure 5 is an example diagram for explaining the script editing process of the script editing module of Figure 3.

[0035] FIG. 6 is a flowchart illustrating a method for generating a context-maintaining image based on a generative artificial intelligence model according to one embodiment of the present invention.

[0036] FIG. 7 is an exemplary diagram showing an image result generated based on a generative artificial intelligence model according to one embodiment of the present invention.

[0037] FIG. 8 is a diagram comparing the similarity between images generated by an image generation method based on a generative artificial intelligence model according to one embodiment of the present invention and an image generation method based on an existing generative artificial intelligence model.

[0038] Figure 9 shows images generated by the present invention and the existing method in the experiment of Figure 8.

[0039]

[0040] Below, embodiments of the present invention are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present invention. The embodiments presented in the present invention are provided to enable those skilled in the art to utilize or practice the contents of the present invention. Accordingly, various modifications to the embodiments of the present invention will be apparent to those skilled in the art. That is, the present invention can be implemented in various different forms and is not limited to the embodiments described below.

[0041] Throughout the specification of the present invention, identical or similar drawing numbers refer to identical or similar components. Furthermore, for the purpose of clearly explaining the present invention, drawing numbers for parts in the drawings that are not relevant to the description of the present invention may be omitted.

[0042] The term "or" as used herein is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified herein or the meaning is clear from the context, "X utilizes A or B" should be understood to mean either of the natural inclusive permutations. For example, unless otherwise specified herein or the meaning is clear from the context, "X utilizes A or B" can be interpreted to mean either X utilizes A, X utilizes B, or X utilizes both A and B.

[0043] The term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the related concepts listed.

[0044] The terms "comprises" and / or "comprising" as used herein should be understood to mean the presence of certain features and / or components. However, it should be understood that the terms "comprises" and / or "comprising" do not exclude the presence or addition of one or more other features, other components, and / or combinations thereof.

[0045] Unless otherwise specified in the present invention or unless the context makes it clear that the singular form is indicated, the singular should generally be construed to include "one or more."

[0046] The term "connection" as used in the present invention should be interpreted to include not only cases where components are "directly connected" but also cases where other components are "present" between them and cases where they are "electrically connected" with other components in between them.

[0047] The term "acquisition" as used in the present invention may be understood to refer to generating or receiving data in an on-device form as well as receiving data via a wireless communication network with an external device or system.

[0048] Meanwhile, the term "module" or "unit" used in the present invention can be understood as a term referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. At this time, the "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, as a narrow concept, a "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a processing process implemented through software execution, or a set of instructions for program execution, etc. In addition, as a broad concept, a "module" or "unit" may refer to the computing device itself that constitutes the system, or an application that runs on the computing device, etc. However, since the above-described concept is only an example, the concept of “module” or “part” can be defined in various ways within a range understandable to those skilled in the art based on the contents of the present invention.

[0049] The term "model" used in the present invention can be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model of a processing process to solve a specific problem. For example, a neural network "model" can refer to the entire system implemented as a neural network that has problem-solving capabilities through learning. In this case, the neural network can have problem-solving capabilities by optimizing the parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a set of neural networks that are a combination of multiple neural networks.

[0050] The term "block" used in the present invention can be understood as a set of components classified based on various criteria such as type, function, etc. Therefore, the components classified as a single "block" can be changed in various ways depending on the criteria. For example, a neural network "block" can be understood as a set of neural networks including at least one neural network. In this case, it can be assumed that the neural networks included in the neural network "block" perform specific operations identically. The explanation of the above-mentioned terms is intended to aid in the understanding of the present invention. Therefore, it should be noted that, unless the above-mentioned terms are explicitly described as matters limiting the contents of the present invention, they are not used in the sense of limiting the technical idea of ​​the contents of the present invention.

[0051] The explanation of the aforementioned terms is intended to aid understanding of the present invention. Therefore, unless explicitly stated as limiting the scope of the present invention, it should be noted that the aforementioned terms are not intended to limit the technical concept of the present invention.

[0052]

[0053] FIG. 3 is a block diagram of a computing device according to one embodiment of the present invention.

[0054] Referring to FIG. 3, a computing device (100) according to one embodiment of the present invention may include a processor (110), a memory (120), and a network unit (130). However, FIG. 3 is merely an example, and the computing device (100) may include other components for implementing a computing environment. In addition, only some of the disclosed components may be included in the computing device (100).

[0055] The processor (110) according to one embodiment of the present invention may be understood as a configuration unit including hardware and / or software for performing computing operations. The processor (110) for performing data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASICc), or a field programmable gate array (FPGA). The type of the processor (110) described above is only an example, and thus the type of the processor (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present invention.

[0056] The processor (110) can train a generative AI model that learns language data such as text and provides image results, and can use the trained generative AI model to provide new content such as text, images, music, audio, and video. During the learning process of the generative AI model, the processor (110) can perform operations that express at least one neural network block included in the model.

[0057] Generative AI models, such as GPT-3, StyleGAN, CycleGAN, WaveNet, DALL-E, and Neural Style Transfer, can learn patterns and relationships from user-generated content datasets and then use these learned patterns to generate new content. Supervised learning can be used to train generative AI models. In supervised learning, a set of user-generated content and its corresponding labels are provided, and the model learns to generate content similar to and labeled with the same user-generated content.

[0058] In this way, generative AI models can be pre-trained on machine learning models, i.e., massive amounts of data, and can be used for tasks beyond content creation, such as improving the quality of digital images, editing videos, rapidly creating manufacturing prototypes, and augmenting data with virtual datasets.

[0059] Meanwhile, since the learning method, structure, and type of the generative AI model described above are only an example, the generative AI model according to one embodiment of the present invention can be configured in various ways based on the example described above.

[0060] The memory (120) may be understood as a configuration unit including hardware and / or software for storing and managing data processed in the computing device (100). That is, the memory (120) may store any type of data generated or determined by the processor (110) and any type of data received by the network unit (130). For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory, a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. In addition, the memory (120) may also include a database system that controls and manages data in a predetermined system. The type of memory (120) described above is only one example, and thus the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present invention.

[0061] The memory (120) can structure and organize and manage data, combinations of data, and program codes executable by the processor (110) required for the processor (110) to perform operations.

[0062] The network unit (130) may be understood as a component that transmits and receives data through any type of known wired or wireless communication system. For example, the network unit (130) may perform data transmission and reception using a wired or wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WIBRO), 5th generation mobile communication (5G), ultra wide-band, Zigbee, radio frequency (RF) communication, wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth. Since the above-described communication systems are only examples, the wired and wireless communication system for data transmission and reception of the network unit (130) may be applied in various ways other than the above-described examples.

[0063] The network unit (130) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system or any client, etc. In addition, the network unit (130) can transmit data generated through calculations of the processor (110) through wired or wireless communication with any system or any client, etc.

[0064] FIG. 4 is a drawing illustrating a detailed configuration of a computing device according to one embodiment of the present invention, and FIG. 5 is an exemplary diagram illustrating a script editing process of the script editing module of FIG. 3.

[0065] Referring to FIG. 4, the computing device (100) includes, but is not limited to, a text input module (210), a script editing module (220), and an image output UI (240).

[0066] The text input module (210) provides a user interface for text-to-image generation to a user terminal, and a script including at least one sentence based on input from the user terminal is input.

[0067] The script editing module (220) separates the script into sentence units, extracts individual sentences, extracts context information from the individual sentences, integrates the context information into the individual sentences based on the context before and after each sentence to form a context sequence, and then edits the script through sentence-level matching using the context sequence.

[0068] The script editing module (220) inputs the edited script as a prompt for the generative AI model (230), and the generative AI model (230) generates an image in which the context between sentences is maintained based on the prompt.

[0069] Previously, since scripts were inputted as text units without distinction of sentences (S) to the generative AI model (230), there was a drawback in that when a script containing various sentences was inputted to the generative AI model (230), only one image was generated for the entire sentence. To solve this, even if the script was inputted into the generative AI model (230) as sentences, the context was not consistently connected between each sentence.

[0070] As shown in FIG. 5, the script editing module (220) creates a script {Text = (S1, S2, .., S n )} as individual sentences (S1, S2, .., S) in the sentence (S) unit n ) and separate the context information (C1, C2, .., C) from individual sentences. n) is extracted, the context information is integrated into individual sentences to create a context sequence (S'1, S'2, .., S' n ) is configured. The script editing module (220) can edit the script by performing sentence-by-sentence matching using the last configured context sequence, and the script edited in this way can be provided as a prompt, which is an input value of the generative AI model (230).

[0071] The image providing module (240) provides an image generated through a generative AI model (230) to a user terminal.

[0072] The modules described above are merely examples for illustrating the present invention, and are not limited thereto and may be implemented in various variations. Furthermore, the modules described above are stored in memory (120) as computer-readable recording media that can be controlled by a processor (110). Furthermore, at least some of the modules described above may be implemented as software, firmware, hardware, or a combination of at least two or more thereof, and may include modules, programs, routines, instruction sets, or processes for performing one or more functions.

[0073] FIG. 6 is a flowchart illustrating an image generation method based on a generative artificial intelligence model according to one embodiment of the present invention, and FIG. 7 is an exemplary diagram showing an image result generated based on a generative artificial intelligence model according to one embodiment of the present invention.

[0074] Referring to FIG. 6, when a script based on input from a user terminal is input (S1), the computing device (100) separates the script into sentence units and extracts individual sentences (S2).

[0075] A sentence (S) is a word (W) that contains a context cxt ) and words that do not contain context (W non ) can be expressed as mathematical formula 1, which is the union of the two sets.

[0076]

[0077]

[0078] The computing device (100) analyzes each separated sentence to obtain context information (W) as in mathematical expression 2, centered on the subject word and the verb word. cxt ) can be extracted (S3).

[0079]

[0080] Here, S cxt is a sentence entered by the user and context information (W cxt ) as a context sequence (or tuple) that is automatically generated from user input sentences and can be used for script editing.

[0081] The computing device (100) integrates context information based on the context before and after each sentence into individual sentences to form a context sequence (S4), and edits the script through sentence-by-sentence matching using the context sequence (S5).

[0082] Specifically, the computing device (100) inputs a user's sentence (S input ) is separated into individual sentences using the same format as mathematical expression 3, and context information is integrated into each sentence to create a context sequence (S n cxt ) can be configured.

[0083]

[0084] The computing device (100) may use an Initial Context First (ICF) method to construct a context sequence.

[0085] The ICF method is based on the observation that ideas or contexts transition within a paragraph. It utilizes the tendency for a paragraph to consist of about five sentences, and for sentences to be short, the main context is often present in the first sentence, and the main context tends to influence the entire paragraph.

[0086] Therefore, the computing device (100) provides context information (W) of the first sentence (S1) of the script. i cxt ) can be duplicated and the duplicated context information can be added to all sentences within the paragraph to form a context sequence as in Equation 4.

[0087]

[0088] The computing device (100) may use Region Context First (RCF) as another method of configuring the context sequence.

[0089] The ICF method may feel like the context in the latter half is disconnected when there are many user input sentences. Therefore, the RCF method uses a sentence threshold that specifies the range in which context continuity should be maintained, and when the sentence threshold is exceeded, new context information (W) is added. i+α cxt ) is extracted to reconstruct the context sequence as in mathematical expression 5. At this time, the sentence threshold (δ) can be adjusted within the range of 5 to 8 sentences to reflect the general structure of the paragraph.

[0090]

[0091] The computing device (100) can selectively use the ICF method and the RCF method as another method for constructing a context sequence. That is, the computing device (100) sets a sentence threshold based on the RCF method, and when a second sentence consecutive to the first sentence is input within a preset time frame after a first sentence in a script is input, if the second sentence is within the sentence threshold, the computing device automatically adds context information of the first sentence to the second sentence based on the ICF method, but if the second sentence exceeds the sentence threshold, the computing device extracts new context information based on the RCF method to reconstruct the context sequence.

[0092] The computing device (100) provides the edited script as a prompt to the generative AI model (230), and the generative AI model (230) can generate an image in which the context between sentences is maintained based on the prompt (S6).

[0093] In Fig. 7, if the script (T) based on user input is T={A man is taking a walk. The man picked up a bag on the road. And then sat down next to the bag}, when the script input by the user is edited to reflect contextual information and the edited script is input as a prompt for the generative AI model, it can be confirmed that the continuous sentence images and animations generated by the generative AI model are images with similar contextual relationships to those generated by the conventional technology in Fig. 2.

[0094] FIG. 8 is a graph comparing the similarity between images generated by an image generation method based on a generative artificial intelligence model according to one embodiment of the present invention and an image generation method based on an existing generative artificial intelligence model, and FIG. 9 shows images generated by the present invention and the existing method in the experiment of FIG. 8.

[0095] The experiment in Figure 8 assumes that context preservation is relatively effective when the similarity between images generated from preceding and following sentences is high, and quantifies the similarity between images generated by the initial context-first method (ICF) and conventional techniques (DALL-E, Stable Diffusion, and MidJourney). For texts consisting of 3 to 5 sentences, an index was created using a vector similarity search algorithm developed by Facebook (Faiss's HNSW) for the generated images, and the similarity was measured and evaluated using the index results.

[0096] In Fig. 8, blue represents the similarity between the image generated from the first sentence and the image generated from the second sentence, orange represents the similarity between the image generated from the second sentence and the image generated from the third sentence, and gray represents the average of all similarities. It was found that the initial context-first method (ICF) of the present invention has higher similarity between images than the conventional method.

[0097] In the present invention, another method of constructing a context sequence, RCF, and a third method that selectively uses the ICF method and the RCF method also showed results similar to ICF.

[0098] In this way, the present invention automatically modifies the prompts to maintain the context of the user input sentences when the user inputs a script consisting of multiple sentences as a prompt for a generative AI model, using a script editing module designed to maintain the context of the user input sentences. Rather than directly modifying the generative AI model itself, the present invention automatically modifies the prompts of the generative AU model to preserve the context.

[0099]

[0100] The embodiments of the present invention described above may also be implemented in the form of a recording medium including computer-executable instructions, such as program modules executed by a computer. Such recording medium includes computer-readable media, and computer-readable media can be any available media that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. Furthermore, computer-readable media includes computer storage media, and computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0101] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0102] The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

Claims

1. A method for generating a context-preserving image based on a generative artificial intelligence model, performed by a computing device including at least one processor, When a script based on user input is input, a step of dividing the script into sentence units and extracting individual sentences; A step of extracting contextual information from the above individual sentences; A step of integrating the context information into individual sentences based on the pre- and post-context of each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence; and A method comprising the step of providing the above-mentioned edited script as a prompt to a generative artificial intelligence model, and generating an image in which the context between sentences is maintained based on the prompt.

2. In paragraph 1, The step of extracting contextual information from the above individual sentences is: A method of extracting contextual information centered on the subject and verb words through morphological analysis in each sentence.

3. In paragraph 1, Based on the context before and after each sentence, the step of integrating the context information into each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence is as follows. A method of duplicating the context information of the first sentence separated from the above script and adding it to all sentences within the paragraph using the Initial Context First (ICF) method.

4. In paragraph 1, Based on the context before and after each sentence, the step of integrating the context information into each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence is as follows. A method of using a Region Context First (RCF) method to set a sentence threshold that specifies a range in which context continuity must be maintained, and extracting new context information to reconstruct a context sequence when the sentence threshold is exceeded.

5. In paragraph 1, Based on the context before and after each sentence, the step of integrating the context information into each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence is as follows. A method comprising a step of adding the context information by selectively using an Initial Context First (ICF) method and a Region Context First (RCF) method.

6. In paragraph 5, The step of adding the context information by selectively using the Initial Context First (ICF) method and the Region Context First (RCF) method is as follows: A step of setting a sentence threshold that specifies the range within which contextual continuity must be maintained; When a first sentence in the above script is input, and a second sentence consecutive to the first sentence is input within a preset time frame, a step of automatically adding context information of the first sentence to the second sentence if the second sentence is within the sentence threshold value; and A method comprising a step of extracting new context information and reconstructing a context sequence when the second sentence exceeds the sentence threshold.

7. A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for generating an image based on a generative artificial intelligence model. The above actions are, When a script based on user input is entered, an action is taken to separate the script into sentence units and extract individual sentences; The act of extracting contextual information from the individual sentences above; An operation of integrating the context information into individual sentences based on the pre- and post-context of each sentence to form a context sequence, and then editing the script through sentence-by-sentence matching using the context sequence; and A computer program comprising an action of providing the above-mentioned edited script as a prompt to a generative artificial intelligence model and generating an image with the inter-sentence context maintained based on the prompt.

Images