Program, information processing device, and information processing method

Abstract

[Problem] To enhance the convenience of image generation in which an image generation model is used. [Solution] Provided is a program for causing a computer to function as a control unit that performs: a process for generating an image of a person using an image generation model constructed by learning a captured image of the person; a process for comparing the person in the generated image with the image of the person and assessing the similarity between said persons; and a process for generating, on the basis of the result of assessing the similarity, input information to be inputted to the image generation model for the purpose of generating an image again.

Description

Program, information processing device, and information processing method

[0001] The present disclosure relates to a program, an information processing device, and an information processing method.

[0002] In recent years, there has been progress in the development of image generation AI (Artificial Intelligence) technology, which automatically generates images from input text. Image generation AI uses an AI model that has learned from many images.

[0003] Since it is difficult to manually prepare a large number of images as training data to be used in machine learning, for example, Patent Document 1 below discloses a technique for generating training data using a CG (Computer Graphics) model.

[0004] International Publication No. 2021 / 177324

[0005] Here, image generation using image generation AI makes it possible to generate a large number of AI images. However, when generating AI images of a specific person, it takes a huge amount of time and effort to visually determine whether the person in each of the generated AI images resembles the person, and then adjust the input information to the image generation AI to obtain the desired AI image.

[0006] Therefore, the present disclosure proposes a program, an information processing device, and an information processing method that can improve the convenience of image generation using an image generation model.

[0007] According to the present disclosure, a program is provided that causes a computer to function as a control unit that performs the following processes: generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate another image based on the results of the similarity determination.

[0008] Furthermore, according to the present disclosure, there is provided an information processing device including a control unit that performs the following processes: generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate an image again based on the result of the similarity determination.

[0009] Furthermore, according to the present disclosure, there is provided an information processing method including: a processor generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person to make a similarity determination; and generating input information for the image generation model to generate the image again based on the result of the similarity determination.

[0010] FIG. 1 is a diagram illustrating an example of the configuration of an information processing system 1 according to an embodiment of the present disclosure. FIG. 2 is a block diagram illustrating an example of the configuration of a server 20 according to this embodiment. FIG. 3 is a diagram for explaining the configuration of an AI model installed in an image generation AI according to this embodiment. FIG. 4 is a flowchart illustrating an example of the flow of an AI advertisement image generation process according to this embodiment. FIG. 5 is a diagram for explaining the generation of an AI image according to LoRA strength and a prompt according to this embodiment. FIG. 6 is a diagram illustrating an example of an AI advertisement image generation screen according to this embodiment.

[0011] Preferred embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0012] The explanation will be given in the following order: 1. Overview 2. Configuration of the server 20 3. Operational processing 4. Display screen example 5. Modified example 6. Supplementary information

[0013] 1 is a diagram illustrating an example of the configuration of an information processing system 1 according to an embodiment of the present disclosure. As shown in Fig. 1, the information processing system 1 includes a model terminal 10 used by fashion models or the like, a model agency terminal 12 used by model agencies or the like to which the fashion models or the like belong, a server 20 (an example of an information processing device), and an advertising agent terminal 30 used by advertising agents.

[0014] The server 20, the model terminal 10, the model office terminal 12, and the advertising agent terminal 30 are connected to each other via a network 40, and can transmit and receive data to and from each other.

[0015] The server 20 generates and outputs advertising images of models using an image generation AI (image generation model). In other words, the server 20 can provide a service for generating AI advertising images. The server 20 can also build an image generation AI using machine learning. The server 20 may build an image generation AI that automatically generates images of specific people by studying captured images (still images, videos) of fashion models or the like sent from the model terminal 10 or the model agency terminal 12. The server 20 generates and provides advertising images of specific people using the image generation AI in response to a request from, for example, the advertising agent terminal 30. It is expected that such advertising images (also referred to as AI advertising images) will be used as advertising materials as they are, or that they will be used to confirm and share the finished image among relevant parties before the actual advertising shoot. The advertising images may be still images or videos.

[0016] The model terminal 10, the model office terminal 12, and the advertising agent terminal 30 are realized by a smartphone, a tablet terminal, a PC (personal computer), or the like.

[0017] A person in charge of a model agency can send captured images of their affiliated models from the model agency terminal 12 to the server 20. In addition to affiliated models, models working independently can also send their captured images to the server 20 from the model terminal 10. By sending captured images of a model to the server 20 and registering the model on the server 20 together with information such as the model's name, age, and height, the model can be presented as a candidate for selection when an advertiser requests the server 20 to generate an AI advertising image.

[0018] The advertising agent terminal 30 requests the server 20 to generate an AI advertising image and obtains the AI ​​advertising image from the server 20. An operation screen for making the generation request may be displayed on the advertising agent terminal 30. The display of such operation screen may be controlled by an application installed on the advertising agent terminal 30, or may be provided by the server 20 and displayed on a web browser.

[0019] The overview of the information processing system 1 according to this embodiment has been described above. In this embodiment, an image generated by the image generation AI is referred to as an AI image, and if the generated image is an advertising image, it is also referred to as an AI advertising image. An advertising image is an image for advertising a specific product or service. Information about the advertising target, a catch phrase, etc. may be included in the prompt input to the image generation AI. Furthermore, the term "image" as used in this specification includes still images and videos.

[0020] Furthermore, the people who are registered as models in the server 20 are not limited to fashion models, but may also be influencers, celebrities, entertainers, athletes, etc.

[0021] The system configuration shown in FIG. 1 is an example, and the present embodiment is not limited thereto. For example, the system configuration shown in FIG. 1 assumes a case where the model agency and the advertising business are separate businesses. However, this is not limited to this, and a business that has the functions of both a model agency and an advertising business, such as a business that registers models and undertakes the production of advertisements using the registered models, is also assumed. In this case, an information processing terminal that integrates the model agency terminal 12 and the advertising business terminal 30 is assumed. In response to a request from an advertiser (company or individual), the business can obtain an AI advertising image using a model registered in the company from the server 20 and provide it to the advertiser.

[0022] 2. Configuration of the Server 20> Fig. 2 is a block diagram showing an example of the configuration of the server 20 according to this embodiment. As shown in Fig. 2, the server 20 includes a communication unit 210, a control unit 220, and a storage unit 230.

[0023] (Communication Unit 210) The communication unit 210 has a transmission unit that transmits data to an external device and a reception unit that receives data from an external device. The communication unit 210 according to this embodiment may be communicatively connected to an external device or the Internet using, for example, a wired or wireless LAN (Local Area Network), Wi-Fi (registered trademark), Bluetooth (registered trademark), a mobile communication network (LTE (Long Term Evolution), 4G (fourth generation mobile communication system), 5G (fifth generation mobile communication system)), or the like.

[0024] (Control Unit 220) The control unit 220 functions as an arithmetic processing unit and a control device, and controls the overall operation of the server 20 in accordance with various programs. The control unit 220 is realized by an electronic circuit such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a microprocessor. The control unit 220 may also include a ROM (Read Only Memory) that stores programs to be used, calculation parameters, etc., and a RAM (Random Access Memory) that temporarily stores parameters that change as appropriate.

[0025] The control unit 220 according to this embodiment can also function as a display control unit 221 , an image generation processing unit 222 , a determination processing unit 223 , and a distribution control unit 224 .

[0026] The display control unit 221 controls the display of various screens, such as an AI advertising image generation request screen and an AI advertising image display screen, which are displayed on the advertising agent terminal 30. Specifically, the display control unit 221 controls the transmission of information required for displaying various screens from the communication unit 210 to the advertising agent terminal 30 via the network 40. On the AI ​​advertising image generation request screen, for example, selection of a model to be used, input of the strength of the AI ​​model, and input of a prompt to be input to the image generation AI can be performed.

[0027] The image generation processing unit 222 uses an image generation AI to generate an AI advertising image of a specified person. The image generation AI may be generated in advance and stored in the image generation AI storage 231. The image generation AI may be generated by the control unit 220 of the server 20. The image generation AI is equipped with an AI model constructed by machine learning predetermined training data (labeled dataset). In this embodiment, an AI model (a so-called text-to-image model) that generates images based on input text is assumed. The control unit 220 may construct the image generation AI using an existing text-to-image model as a foundation, by additionally learning the characteristics of a specific person (e.g., a labeled dataset obtained by analyzing captured images) from the foundation AI model (so-called fine tuning). The dataset may be obtained from still images or videos of a specific person.

[0028] FIG. 3 is a diagram illustrating the configuration of the AI ​​model installed in the image generation AI according to this embodiment. As shown in FIG. 3, a Foundation Model 510, which is a basic AI model (text-to-image model), is loaded with a generic Japanese model 520 obtained through additional training and a proper noun model 530 trained on captured images of specific fashion models, etc. The generic Japanese model 520 is obtained by additional training using captured images of, for example, several thousand Japanese people. Note that while the generic Japanese model 520 is generated here as an example, the control unit 220 may also generate a generic Indian model or a generic American model. The proper noun model 530 is obtained by additional training using captured images of specific fashion models, etc., transmitted from the model terminal 10 or the model agency terminal 12. Low Rank Adaptation (LoRA) may be used for this additional training. The proper noun model 530 is also referred to as a specific model LoRA. The proper noun model 530 can be constructed for each specific person (such as a registered fashion model).

[0029] When performing the process of generating an AI advertising image using image generation AI, the image generation processing unit 222 generates an AI advertising image of a specific person using combination a of the Foundation Model 510, the generic Japanese model 520, and the proper noun model 530 shown in FIG. 3 , or combination b of the Foundation Model 510 and the proper noun model 530. The proper noun model 530 used is assumed to be an AI model corresponding to the specified person (i.e., an AI model that has learned captured images of the specified person). This allows the image generation processing unit 222 to generate an AI advertising image of the specified person.

[0030] The configuration of the AI ​​model installed in the image generation AI described above is an example, and the present embodiment is not limited to this.

[0031] The image generation processing unit 222 also generates (acquires) information to be input to the image generation AI. The input information may be information input by the advertiser (user) and transmitted from the advertising agent terminal 30, or may be information appropriately generated by the image generation processing unit 222. The image generation processing unit 222 also generates (modifies) input information to be input to the image generation AI to generate an AI advertising image again, depending on the determination result by the determination processing unit 223 described below. For example, the image generation processing unit 222 may randomly change the input numerical values.

[0032] Examples of information to be input to the image generation AI include prompts. Examples of prompts include age, race, hairstyle, hair color, facial features, makeup (makeup intensity, eyebrow density, presence or absence of eyeliner or eyeshadow, eyelash condition, blush color, lipstick color, type of makeup, etc.), body type, pose, etc. The image generation processing unit 222 may generate a prompt from a captured image of a specified model using an image-to-text generation AI.

[0033] Another example of information input to the image generation AI is checkpoints. Checkpoints are information input for use with an AI model that generates images of generic people, such as the generic Japanese model 520. For example, checkboxes for race, gender, etc. are presented to the user, and the image generation processing unit 222 uses an AI model that matches the information (checkpoints) checked by the user.

[0034] The information input to the image generation AI also includes various parameters. The various parameters include, for example, the strength (strength of influence) of the AI ​​model to be used. For example, the strength of the proper noun model 530 (specifically, LoRA strength) can be included.

[0035] The determination processing unit 223 performs a similarity determination between a person in the AI ​​advertising image generated by the image generation processing unit 222 and a specified person. The determination processing unit 223 may compare an image of a person in the AI ​​advertising image with a captured image of the specified person, and perform a similarity determination (determine whether or not they are similar) based on whether the similarity index obtained using a determination AI that outputs a similarity index exceeds a threshold. The determination AI is generated in advance and stored in the determination AI storage 232. The algorithm of the determination AI is not particularly limited, but for example, the AI ​​advertising image and the captured image may be analyzed to extract feature vectors of the person, and the distance and direction between the feature vectors may be compared to evaluate the similarity between the person in the AI ​​advertising image and the specified person. The similarity index may be calculated, for example, as a value between 0 and 1, with the closer to 1 the evaluation, the closer the index is to the comparison target (the specified person).

[0036] The distribution control unit 224 controls the distribution of the AI ​​advertising image generated by the image generation processing unit 222. The distribution control unit 224 may, for example, request a server (not shown) of an advertising distribution company to distribute the AI ​​advertising image. The distribution control unit 224 may control the automatic distribution of the AI ​​advertising image determined by the determination processing unit 223 to be similar to a specified person, or may control the distribution when permitted by the advertiser.

[0037] (Storage Unit 230) The storage unit 230 is realized by a ROM that stores programs and calculation parameters used in the processing of the control unit 220, and a RAM that temporarily stores parameters that change as needed.

[0038] The storage unit 230 stores an image generation AI storage 231, a judgment AI storage 232, a generation information storage unit 233, and a personal image storage unit 234. The image generation AI storage 231 stores image generation AI. The judgment AI storage 232 stores judgment AI. The generation information storage unit 233 stores information related to the generation of AI advertising images. For example, the generation information storage unit 233 stores prompts and the like used to generate the AI ​​advertising images. The generation information storage unit 233 also stores generated AI advertising images. The personal image storage unit 234 stores captured images of fashion models and the like.

[0039] Although the configuration of the server 20 has been specifically described above, the configuration of the server 20 according to the present disclosure is not limited to the example shown in Fig. 2. For example, the server 20 does not necessarily have to have all of the components shown in Fig. 2. Furthermore, the server 20 may be realized by a plurality of devices.

[0040] 3. Operation Processing FIG. 4 is a flowchart showing an example of the flow of the AI ​​advertisement image generation processing according to this embodiment.

[0041] 4, first, the image generation processing unit 222 of the server 20 selects a specific model LoRA based on a user operation of an advertiser or the like (step S103). Specifically, the image generation processing unit 222 selects a specific model LoRA (an example of a proper noun model) that has learned a captured image of a specific person (such as a fashion model) designated by the user.

[0042] Next, the image generation processing unit 222 acquires input information for the image generation AI (step S106). For example, the image generation processing unit 222 acquires a checkpoint, LoRA strength, and a prompt as input information. The input information may be input by a user. The image generation processing unit 222 may also generate a prompt from a captured image of a specific person designated by the user. For example, the image generation processing unit 222 may use an image-to-text generation AI to generate a prompt such as "about 25 years old, Japanese, black hair, large eyes, thick eyebrows" from a captured image (studio shot image) of a specific person designated by the user, and add the generated prompt to the input information.

[0043] Next, the image generation processing unit 222 generates an AI advertising image using the image generation AI (step S109). Specifically, the image generation processing unit 222 inputs the acquired input information to the image generation AI and outputs an AI advertising image. It is expected that a large number of AI advertising images will be output from the image generation AI.

[0044] Next, the determination processing unit 223 references the image of the specific person designated by the user and determines the similarity between the generated AI advertising image and the specific person (step S112). Specifically, the determination processing unit 223 calculates a similarity index using the determination AI. The specific person's image is, for example, a captured image taken in a studio or the like, transmitted from the model terminal 10 or the model agency terminal 12 when the model is registered in the server 20 and used for additional learning (model LoRA generation). Such captured image is stored in the image storage unit 234 and can also be used for similarity determination. The determination processing unit 223 compares each of the large number of output AI advertising images with the image of the specific person and calculates a similarity index. In this way, this embodiment makes it possible to automatically determine whether the AI ​​advertising image resembles the specified specific person. When generating AI advertising images using image generation AI, a large number of images are output, so performing automatic determination can reduce the user's effort.

[0045] Next, the determination processing unit 223 determines whether the similarity index for each AI advertising image exceeds a threshold value (step S115). For example, if the similarity index is closer to 1, the person resembles the person, the threshold value is set to 0.8, and the threshold value is used as the criterion for determining whether the person in the AI ​​advertising image resembles a specific person specified by the user. The threshold value shown as an example is merely an example, and the present embodiment is not limited to this.

[0046] Next, if there is an AI advertising image whose similarity index exceeds a threshold value (step S115 / Yes), the control unit 220 decides to adopt the AI ​​advertising image and controls to deliver the advertisement (step S118). Note that while advertisement delivery has been described as an example here, the present invention is not limited to this. The control unit 220 may also present an AI advertising image whose similarity index exceeds a threshold value to the user. Furthermore, the control unit 220 may also present the AI ​​advertising image to the user together with the similarity index and the determination result. Then, the control unit 220 may control to deliver the AI ​​advertising image if permitted by the user.

[0047] On the other hand, if there is no AI advertising image whose similarity index exceeds the threshold (step S115 / No), the image generation processing unit 222 changes the input information to the image generation AI (step S121) and returns to step S109 to generate an AI advertising image again. For example, the image generation processing unit 222 changes the LoRA intensity or the prompt. The image generation processing unit 222 may randomly change the LoRA intensity or the content of the prompt, or may adjust the change of the input information depending on the determination result. While it is desirable to generate input information that outputs an AI advertising image whose similarity index is closer to 1, the image generation processing unit 222 may randomly change the input information and repeat steps S109 to S115 and S121 to obtain an AI advertising image whose similarity index exceeds the threshold. Alternatively, the image generation processing unit 222 may learn the input information and the determination result to generate a prediction model and change the input information using the prediction model.

[0048] FIG. 5 is a diagram for explaining the generation of AI images according to LoRA strength and prompts in this embodiment. Here, the LoRA strength is a setting for the strength of the influence effect of a specific model LoRA, and as shown in FIG. 5, various settings are possible. A LoRA strength of 1 does not necessarily generate an AI image that most resembles the individual; the image generation processing unit 222 appropriately changes the LoRA strength to generate the image. The generated AI image also has a different impression depending on the prompt setting. While prompts do not necessarily include age, taking age as an example, AI images with different impressions may be generated for ages 20, 24, and 28. Even if the age specified in the prompt is the individual's actual age, the AI ​​image that most resembles the individual may not necessarily be generated; the image generation processing unit 222 appropriately changes the age to generate the image. FIG. 5 shows a simplified example of an AI image generated by an image generation AI using a specific model LoRA corresponding to a specific person specified by the user, according to a combination of LoRA strength and prompt.

[0049] For example, in the example shown in Figure 5, among the AI ​​images generated when the LoRA strength is 0.8, 0.9, and the prompt (age): 28 years old, the similarity index may exceed the threshold when compared with the image of the person in AI image 600 with an LoRA strength of 0.9 and an age of 28 years old.

[0050] Although the flow of the AI ​​advertising image generation process according to this embodiment has been described above, this embodiment is not limited to this. For example, in step S115, steps S109 to S115 and S121 may be repeated until a predetermined number of AI advertising images whose similarity index exceeds a threshold are generated. Then, when a predetermined number of AI advertising images whose similarity index exceeds a threshold are generated, all of the AI ​​advertising images may be presented to the user as advertising candidates.

[0051] 4. Display Screen Example Next, a display screen example for generating an AI advertising image according to this embodiment will be described with reference to the drawings. Here, the display screen displayed on the advertising agent terminal 30 will be described as an example.

[0052] 6 is a diagram showing an example of an AI advertisement image generation screen according to this embodiment. As shown in FIG. 6, an AI advertisement image generation screen 700 displays a model LoRA selection screen 710, a checkpoint input screen 720, a LoRA strength input screen 730, a prompt input screen 740, a model person image display screen 750, a generate button 701, and a result display screen 760.

[0053] On the model LoRA selection screen 710, a proper noun model 530 to be used in combination with the Foundation Model 510 is selected (input). Specifically, on the model LoRA selection screen 710, a model LoRA corresponding to a specific person designated by the user is selected. The user may select from each LoRA file in a pull-down format, or the user may select a person to use from a model list screen (not shown), and the server 20 may automatically select a corresponding model LoRA in response and display it on the model LoRA selection screen 710.

[0054] In the checkpoint input screen 720, a generic model to be used in combination with the Foundation Model 510 is selected (input). Here, as an example, "Japanese woman.file", which is the generic Japanese model 520, is input. The generic model may be selected by the user in a pull-down menu. Alternatively, the user may check race, gender, etc. on a checkbox screen (not shown), and the server 20 may automatically select the corresponding generic model and display it on the checkpoint input screen 720.

[0055] The user can arbitrarily input the LoRA strength on the LoRA strength input screen 730. The LoRA strength may be input in a format having a fixed range.

[0056] The prompt input screen 740 allows the user to input a prompt at will. Examples of the input prompt include facial expression, clothing, hairstyle, age, makeup, target of advertisement (such as product ID), pose, atmosphere of the advertisement, scene (location, environment, background), composition, catch phrase, etc. The product ID is identification information of product information registered in a product database (not shown). The product database is stored in, for example, the server 20. Product information (such as product images) can be registered by the user from, for example, the advertiser terminal 30.

[0057] The prompt does not necessarily have to be input by the user, but may be automatically generated by the server 20 from a captured image of a specific person (for example, an image taken in a studio) and displayed on the prompt input screen 740. The user can check the automatically generated prompt and make corrections or additions.

[0058] On the model's own image display screen 750, a captured image of a fashion model or the like registered on the server 20 is displayed.

[0059] The generate button 701 is selected when generating an AI advertising image. When the generate button 701 is selected, the server 20 inputs input information to the image generation AI, and generates (outputs) an AI advertising image.

[0060] The result display screen 760 displays the generated AI advertising images and the judgment results (including the similarity index) of each AI advertising image. Here, as an example, AI advertising images whose similarity index exceeds a threshold, i.e., AI advertising images judged to be similar to the model, and AI advertising images whose similarity index does not exceed the threshold, i.e., AI advertising images judged not to be similar to the model, are displayed. When checking the generated AI advertising images, the user can refer to the judgment results by the system. Note that the result display screen 760 may be controlled to display AI advertising images whose similarity index is equal to or greater than a threshold (e.g., 0.5) lower than the threshold (e.g., 0.8), for example.

[0061] The user may refer to the generated AI advertising image and the determination result by the system to modify the prompt, etc., and select the generate button 701 again to have the server 20 repeatedly generate the AI ​​advertising image. It is also possible to have the server 20 perform all of the modifications to the input information. For example, although not shown in the figure on the AI ​​advertising image generation screen 700 in FIG. 6, an "auto-adjust" checkbox may be provided, and when the generate button 701 is selected while the checkbox is checked, the server 20 may change the input information and then generate the AI ​​advertising image again.

[0062] <5. Modifications> When determining the similarity between a person in an AI advertising image and the model himself (a specific person designated by a user), in the above-described embodiment, an image taken at the time of model registration (for example, an image taken in a studio) is referenced as the image of the person himself. However, if time has passed since the model was registered, the model may have aged and his or her overall appearance, such as face, figure, hairstyle, etc., may have changed.

[0063] Therefore, the server 20 may acquire a captured image showing the current appearance of the model as appropriate and use it as the model's image to be referenced in the determination process. After a certain period of time has passed, the server 20 may request the registered model or the agency to send a captured image showing the current appearance, or may collect captured images showing the current appearance from the SNS (Social Networking Service) of the registered model or the agency.

[0064] In addition, when the server 20 automatically generates a prompt to be input to the image generation AI from the model's own image, it may use an image of the model's own image that was taken at a date and time closer to the present.

[0065] <6. Supplementary Information> Although preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings, the present technology is not limited to such examples. It is clear that a person skilled in the art of the present disclosure can conceive of various modified or altered examples within the scope of the technical idea described in the claims, and it is understood that these also naturally fall within the technical scope of the present disclosure.

[0066] For example, in the above-described embodiment, an advertising AI image was given as an example of an AI image generated using image generation AI, but this embodiment is not limited to this and may be applied to purposes other than the generation of advertising AI images.

[0067] It is also possible to create one or more computer programs for causing the hardware, such as the CPU, ROM, and RAM, built into the server 20 to perform the functions of the server 20. A computer-readable storage medium storing the one or more computer programs is also provided.

[0068] Furthermore, the effects described herein are merely descriptive or exemplary and are not limiting. In other words, the technology according to the present disclosure may achieve other effects that will be apparent to those skilled in the art from the description of this specification, in addition to or in place of the above-described effects.

[0069] The present technology can also be configured as follows. (1) A program that causes a computer to function as a control unit that performs the following processes: generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and performing a similarity determination; and generating input information for the image generation model to generate an image again based on the result of the similarity determination. (2) The program described in (1), in which the control unit determines whether images are similar based on whether a similarity index calculated in the similarity determination exceeds a threshold. (3) The program described in (1) or (2), in which the control unit changes the input information for the image generation model and controls the generation of an image of the person again using the image generation model when it is determined that, among images generated using the image generation model, there is no image whose similarity index calculated in the similarity determination exceeds a threshold. (4) The program described in any one of (1) to (3), in which the input information input to the image generation model includes a prompt. (5) The program according to any one of (1) to (4), wherein the input information input to the image generation model includes AI model strength. (6) The program according to any one of (1) to (5), wherein the control unit uses a captured image of the person used when constructing the image generation model as the image of the person to reference when making the similarity determination. (7) The program according to any one of (1) to (5), wherein the control unit uses a captured image with a capture date and time closer to the present as the image of the person to reference when making the similarity determination. (8) The program according to any one of (1) to (7), wherein the control unit generates a prompt from the captured image of the person and uses the prompt as input information to be input to the image generation model. (9) The program according to any one of (1) to (8), wherein the control unit randomly changes the input information when generating an image again. (10) The program according to any one of (1) to (8), wherein the control unit, when generating an image again, changes the input information using a prediction model that has learned the input information and the result of similarity determination.(11) The program according to any one of (1) to (10), wherein the control unit additionally trains a trained AI model with a dataset generated from captured images of the person, thereby constructing an image generation AI as the image generation model. (12) The program according to any one of (1) to (11), wherein the control unit generates an advertisement image as the image of the person. (13) The program according to any one of (12), wherein the control unit controls delivery of the advertisement image determined to be similar to the person by similarity determination. (14) The program according to any one of (1) to (13), wherein the control unit performs processing to present the generated image of the person and the result of the similarity determination to a user. (15) An information processing device comprising a control unit that performs processing to generate an image of the person using an image generation model constructed by learning captured images of the person, processing to compare the person in the generated image with the image of the person and perform similarity determination, and processing to generate input information for the image generation model to generate an image again, based on the result of the similarity determination. (16) An information processing method including: a processor generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate another image based on the result of the similarity determination.

[0070] REFERENCE SIGNS LIST 1 Information processing system 10 Model terminal 12 Model office terminal 20 Server 210 Communication unit 220 Control unit 221 Display control unit 222 Image generation processing unit 223 Determination processing unit 224 Distribution control unit 230 Storage unit 231 Image generation AI storage 232 Determination AI storage 233 Generation information storage unit 234 Personal image storage unit 30 Advertiser terminal 40 Network

Claims

1. A program that causes a computer to function as a control unit that performs the following processes: generating an image of a person using an image generation model constructed by learning from captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate another image based on the results of the similarity determination.

2. The program according to claim 1, wherein the control unit determines whether the images are similar based on whether the similarity index calculated in the similarity determination exceeds a threshold value.

3. The program described in claim 1, wherein, when it is determined that there is no image among the images generated using the image generation model whose similarity index calculated in the similarity judgment exceeds a threshold, the control unit changes the input information to the image generation model and controls the generation of an image of the person again using the image generation model.

4. The program of claim 1, wherein the input information input to the image generation model includes a prompt.

5. The program of claim 1, wherein the input information input to the image generation model includes AI model strength.

6. The program according to claim 1, wherein the control unit uses the captured image of the person used when constructing the image generation model as the image of the person to reference when making the similarity determination.

7. The program according to claim 1, wherein the control unit uses an image captured at a date and time closest to the present as the image of the person to be referenced when making the similarity determination.

8. The program according to claim 1, wherein the control unit generates a prompt from a captured image of the person and uses the prompt as input information to be input to the image generation model.

9. The program according to claim 1, wherein the control unit randomly changes the input information when generating an image again.

10. The program according to claim 1, wherein the control unit, when generating an image again, modifies the input information using a predictive model that has learned the input information and the results of similarity determination.

11. The program described in claim 1, wherein the control unit constructs an image generation AI as the image generation model by additionally training a trained AI model with a dataset generated from captured images of the person.

12. The program according to claim 1, wherein the control unit generates an advertising image as the image of the person.

13. The program according to claim 12, wherein the control unit controls the distribution of the advertising image that is determined to resemble the person by the similarity determination.

14. The program according to claim 1, wherein the control unit performs processing to present the generated image of the person and the results of the similarity determination to a user.

15. An information processing device comprising a control unit that performs the following processes: generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate another image based on the results of the similarity determination.

16. An information processing method comprising: a processor generating an image of a person using an image generation model constructed by learning captured images of the person; comparing the person in the generated image with the image of the person and making a similarity determination; and generating input information for the image generation model to generate another image based on the result of the similarity determination.

Images