Content presentation system, content presentation method, and program

Abstract

To effectively present images generated by image generation AI to improve the sense of immersion while reading.SOLUTION: A content presentation system comprises: a text information acquisition unit that acquires text information; a scene generation prompt generating unit that generates, from the text information, a scene generation prompt for drawing a scene of a text corresponding to the text information; a scene image generation unit that generates a scene image, which is a still image or a moving image, corresponding to the text information by inputting the scene generation prompt into an image generation model; and a content presenting unit that presents the scene image together with the text information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a content presentation system, a content presentation method, and a program. [Background technology]

[0002] There is generative AI technology that uses a model trained on large amounts of data to generate new data with the same characteristics as the training data. One type of generative AI technology is image generation AI, which generates new images by learning from large amounts of images. Image generation AI can be broadly categorized into three types of image generation models: VAE (Variational Auto Encoder), GAN (Generative Adversarial Network), and diffusion model. Of the image generation models, diffusion models generate high-quality images, capable of generating images that are indistinguishable from real photographs or human-made drawings and illustrations. Furthermore, diffusion models commonly use a method called Text-to-Image, in which a string of characters (prompt) specifying the image to be generated is input and an image is output.

[0003] With the development of image generation technology such as diffusion models and signal processing technology, it has become possible to generate images from character strings at high speed and with high accuracy, and content utilizing image generation AI has emerged. For example, Non-Patent Document 1 discloses technology that uses image generation AI to automatically generate illustrations according to the content of a novel. Non-Patent Document 2 discloses an example of a novel. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Ryusei Hisanaga and Taketoshi Ushiami, "Automatic Generation of Novel Illustrations Using Prompt Programming," 15th Forum on Data Engineering and Information Management, March 2023 [Non-patent document 2] "The Little Match Girl" by Hans Christian Andersen, translated by Okubo Yu, [online], Aozora Bunko, March 27, 2021, [Retrieved May 27, 2024], Internet<https: / / www.aozora.gr.jp / cards / 000019 / files / 194_23024.html> Summary of the Invention [Problem to be solved by the invention]

[0005] The technology described in Non-Patent Document 1 uses a diffusion model specialized for Japanese (Japanese-Stable-Diffusion) as an image generation AI. Because the diffusion model generates images using random numbers, it does not necessarily generate illustrations that match the reader's imagination. Furthermore, the technology described in Non-Patent Document 1 extracts information indicating who is doing what as scene features and inputs the extracted scene features into the image generation AI to generate images. This generates images that include the main characters or characters in a novel scene. In this case, images of the same person or character may be generated with different appearances on different pages, resulting in a lack of consistency between pages. When such inconsistent images are presented as illustrations, people may be distracted by the images, their attention to the novel may be distracted, and the sense of immersion during reading may be impaired.

[0006] The present invention was made in consideration of this situation, and its purpose is to provide a content presentation system, content presentation method, and program that can effectively present images generated by image generation AI and improve the sense of immersion in reading. [Means for solving the problem]

[0007] One aspect of the present invention is a content presentation system comprising a text information acquisition unit that acquires text information, a scene generation prompt generation unit that generates a scene generation prompt from the text information for drawing a scene of a text corresponding to the text information, a scene image generation unit that generates a scene image, which is a still image or a moving image, corresponding to the text information by inputting the scene generation prompt into an image generation model, and a content presentation unit that presents the scene image together with the text information.

[0008] One aspect of the present invention is a content presentation method performed by a computer, in which a text information acquisition unit acquires text information, a scene generation prompt generation unit generates a scene generation prompt from the text information for drawing a scene of a text corresponding to the text information, a scene image generation unit generates a scene image, which is a still image or a moving image, corresponding to the text information by inputting the scene generation prompt into an image generation model, and a content presentation unit presents the scene image together with the text information.

[0009] One aspect of the present invention is a program that causes a computer to acquire text information, generate from the text information a scene generation prompt for drawing a scene of a text corresponding to the text information, input the scene generation prompt into an image generation model to generate a scene image that is a still image or a moving image corresponding to the text information, and present the scene image together with the text information. [Effects of the Invention]

[0010] According to the present invention, images generated by image generation AI can be effectively presented, improving the sense of immersion in reading. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a block diagram illustrating an example of a configuration of a content presentation system according to an embodiment. [Figure 2] 10A and 10B are diagrams for explaining a process performed by a prompt generation unit for generating a scene according to an embodiment. [Figure 3] 3A to 3C are diagrams for explaining processing performed by a scenery image generating unit of the embodiment. [Figure 4] 10A and 10B are diagrams illustrating examples of images displayed by a content presenting unit according to an embodiment. [Figure 5] 10 is a flowchart illustrating a flow of processing performed by the content presentation system according to the embodiment. [Figure 6] 10A and 10B are diagrams illustrating an example in which a content presenter of the embodiment displays an image on a tablet terminal. [Figure 7] 10A and 10B are diagrams illustrating an example in which a content presenter of the embodiment projects an image. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0013] (Regarding the content presentation system 10) The content presentation system 10 is a system that presents images that correspond to the text written on each page of a book as content. 1 is a block diagram showing an example of the configuration of a content presentation system according to an embodiment. The content presentation system 10 includes a text information acquisition unit 101, a scene generation prompt generation unit 102, a scene image generation unit 103, and a content presentation unit 104. The functional units of the content presentation system 10 (the text information acquisition unit 101, the scene generation prompt generation unit 102, the scene image generation unit 103, and the content presentation unit 104) may be provided in a single computer or may be distributed across multiple different computers.

[0014] The text information acquisition unit 101 acquires text information T indicating the text written on each page of a target book that is the target of image generation. For example, the text information acquisition unit 101 acquires text information in text format (text information) from a text file indicating the text written on each page of the target book. The text information may be pre-stored in the content presentation system 10, or may be stored in an external database. The text information acquisition unit 101 accesses a storage unit of the content presentation system 10 or an external database in response to a user operation or the like, and acquires the text information of the target book, thereby acquiring text information T indicating the text written on each page of the target book.

[0015] For example, the text information acquisition unit 101 may acquire the text information T from electronic book data obtained by digitizing the target book, or may acquire the text information T obtained by performing character recognition on images of each page of the target book captured by a camera. Furthermore, the text information acquisition unit 101 may acquire the text information T obtained by performing character recognition on images of the target book read by a scanner. The text information acquisition unit 101 outputs the acquired text information T to the scene generation prompt generation unit 102 .

[0016] The scene generation prompt generation unit 102 acquires text information from the text information acquisition unit 101 and generates a prompt based on the acquired text information. The prompt is information that instructs what kind of image to generate in image generation, and in this case, is information that instructs the generation of an image that expresses a scene shown in a sentence corresponding to the text information.

[0017] In this embodiment, the scene generation prompt generation unit 102 generates a prompt that indicates a scene (a scene generation prompt). The scene here refers to a landscape, background, emotion, concept, image, or the like depicted in a sentence. The scene here excludes people, characters, animals, and the like whose facial expressions and actions are described in a sentence with distinctive features and tend to attract a large amount of attention from the reader. The scene also includes landscapes and the like that tend to attract a small amount of attention from the reader, as it is thought that many readers do not pay much attention to details.

[0018] By generating such a scene generation prompt, an image can be generated that includes only landscapes and other elements that are less likely to draw attention to details, and excludes portraits and other elements that attract the reader's attention. Readers often do not pay much attention to the landscapes depicted in text, but are instead drawn to people and characters, forming their own impressions. In light of this tendency, this embodiment generates an image from content extracted from text that does not include people and characters, such as those whose facial shapes, physiques, facial expressions, hairstyles, clothing, poses, and angles are likely to draw attention, but includes landscapes and other elements whose details are less likely to draw attention. This prevents the generation of an image that does not match the reader's image, thereby preventing the reader from becoming immersed in reading.

[0019] 2 is a diagram illustrating the processing performed by the prompt generation unit for generating a scene according to the embodiment. The prompt generation unit for generating a scene 102 extracts scene features related to a scene from text information T using a large-scale language model, and generates a prompt for generating a scene using the extracted scene features. The large-scale language model is a natural language processing model trained using a large amount of text data. As shown in FIG. 2, in the process of generating scene information J, the scene generation prompt generation unit 102 uses two inputs IA and IB to output scene features J generated by a large-scale language model as an output Out as a scene generation prompt. The input IA is set to text information T. For example, based on a book designated by the user, each time the user turns a page in the book, the text information T written on that page is set to the input IA. A fixed instruction F is input to the input IB. The fixed instruction F is a common instruction fixed to each page of a book, and specifies that a scene feature J should be generated that includes only scenes extracted from the text information T (text "A") set in the input IA, and does not include people, characters, etc. Here, the fixed instruction F may be provided with a setting "B" that specifies the output format. For example, a character string such as "an English prompt of approximately 20 characters" is set in setting "B." By providing a setting that specifies the output format in this way, a prompt suitable for the image generation AI used for image generation can be generated. For example, by instructing to generate a prompt in "English," a prompt can be generated in a format that can be used as input for a general image generation model. Furthermore, by limiting the number of characters to approximately 20 characters or the number of words to 10 to 30, it is possible to prevent a prompt from being too short to extract necessary information, or too long to extract redundant information or information other than the scene. This allows an appropriate output format to be specified in accordance with the development of the image generation AI. Then, the scene generation prompt generation unit 102 inputs a fixed instruction F to the large-scale language model. The large-scale language model generates a scene generation prompt according to the input fixed instruction F. The large-scale language model outputs the generated scene generation prompt, and the scene generation prompt generation unit 102 sets the scene generation prompt generated by the large-scale language model to the output Out. The scene generation prompt generating unit 102 outputs the generated prompt (scene generation prompt) to the scene image generating unit 103.

[0020] Returning to the explanation of FIG. 1, the scene image generation unit 103 acquires a prompt from the scene generation prompt generation unit 102 and generates an image based on the acquired prompt. The scene image generation unit 103 uses, for example, a diffusion model as an image generation AI and generates a scene image JG by inputting a prompt using Text-to-Image of the diffusion model. That is, the scene image generation unit 103 inputs the prompt generated by the scene generation prompt generation unit 102 to the image generation AI. As a result, the scene image generation unit 103 generates an image (scene image JG) that represents a scene shown in a sentence corresponding to the sentence information. In this embodiment, the scenery image may be a still image or a moving image.

[0021] 3 is a diagram for explaining the processing performed by the scene image generation unit 103 of the embodiment. As shown in Fig. 3, in the processing for generating an image, the scene image generation unit 103 sets a prompt PT and inputs the set prompt PT into a diffusion model as an image generation model, thereby generating a scene image JG. The scene image generation unit 103 sets, for example, a positive prompt PPT and a negative prompt NPT as prompts PT. The positive prompt PPT is a prompt that indicates the content that is to be included in an image when the image is generated. The negative prompt NPT is a prompt that indicates the content that is to be excluded from the image to be generated. By inputting elements of both the positive prompt and the negative prompt, it is possible to clarify what is to be included and what is to be excluded from the image, thereby improving the accuracy of the generated image.

[0022] This diagram shows an example in which two positive prompts PPT1 and PPT2 are set. The positive prompt PPT1 is a prompt (a prompt for generating a scene) generated by the prompt generating unit 102 for generating a scene. Positive prompt PPT2 is used to ensure consistency among images corresponding to each page of a book, and is a prompt that indicates the style of the image, for example. By fixing the style, it is possible to unify the consistency and atmosphere between pages of a book to a certain extent, reducing the disruption of immersion. It is also possible to specify a style such as a picture book style or a painting style. By allowing the user to specify their preferred style using a common positive prompt, it is possible to maintain consistency and atmosphere between pages and generate images that suit the user's preferences.

[0023] This figure also shows an example of setting one negative prompt NPT. A negative prompt is a prompt that specifies elements that should not be included in the scenery image JG. The scenery image generator 103 sets elements in the image that are likely to attract the reader's attention, such as humans, characters, and animals (humans, characters, men, women, people, living beings, animals, creatures), as negative prompts NPT. The scenery image generating unit 103 may specify a negative prompt NPT in which elements such as a person or character are set as a standard prompt for all pages.

[0024] In this way, the positive prompt PP1 specifies elements that are difficult to focus on in detail, such as a scene, and the negative prompt NPT specifies the exclusion of elements that are easy to focus on in detail, such as people. This allows for the generation of an image that includes elements that are difficult to attract the reader's attention, but does not include elements that are likely to attract the reader's attention, preventing elements that disrupt the reader's immersion when reading from being reflected in the image. Additionally, the style specification shown in the positive prompt PP2 and the exclusion specification shown in the negative prompt NPT are set as common settings for all pages. This allows for consistency and atmosphere to be maintained between pages. For example, by allowing the user to set their preferred style in advance, it is possible to generate images that match the user's preferences so that each page of a book is consistent.

[0025] Here, in the learning process of image generation AI, a large number of images of people and characters are often learned from a large dataset. Therefore, when an image generation AI is made to generate an image, there is a possibility that the image generated will unintentionally resemble another person's copyrighted work. In a system such as the content presentation system 10 of the present embodiment, which interactively presents an image corresponding to the text on a page each time a user turns a page, the generated image is presented directly to the user, making it difficult to check the generated image in advance. Therefore, if an image similar to another person's copyrighted work is generated, this can become a significant problem. In contrast, in the content presentation system 10 of this embodiment, a negative prompt NPT is set for all pages to specify that images should not include people, characters, animals, or other elements that tend to attract the reader's attention. This makes it easier to generate a scenery image JG that depicts only things that are unlikely to draw attention to details, such as scenery described in text. This makes it harder to generate images that resemble other people's works.

[0026] Furthermore, the scene image generation unit 103 may remove concepts such as humans and characters from the pre-trained image generation model by using an additional learning method such as LECO (Low-rank adaptation for Erasing Concepts from Diffusion Models). By using an image generation model that makes it difficult to generate elements such as humans, characters, and living things, it is possible to prevent the generation of images that are likely to attract people's attention.

[0027] The scenery image generation unit 103 outputs the generated image (scenery image JG) to the content presentation unit 104.

[0028] Returning to the explanation of FIG. 1, the content presenting unit 104 receives as input text information T in text format acquired from the text information acquiring unit 101 and a scene image JG generated by the scene image generating unit 103, and presents them on various devices using computer graphics technology or the like. This allows the reader to appreciate the text information and the scene image simultaneously. That is, the content presenting unit 104 acquires an image from the scene image generating unit 103 and presents the acquired image together with the text information T. As a result, the image generated by the scene image generating unit 103, specifically an image expressing a depiction in which details are not easily noticed, such as a landscape extracted from a sentence corresponding to the text information T, is presented together with the text information T.

[0029] 4 is a diagram illustrating an example of an image displayed by the content presenter of the embodiment. In this diagram, an example of an image (content image) is shown in which text information T is displayed in a rectangular display area R and a scenery image JG is displayed on the periphery surrounding the rectangular display area R.

[0030] As shown in this figure, the content presenter 104 may present, for example, the text information T in a partially overlapping manner in the foreground of the scene image JG. This causes the scene image JG to be displayed in the periphery of the text information T or as the background of the text information T. This results in the scene image JG being presented in the user's peripheral vision or behind the text written in the book, making it possible to present the scene image JG indirectly rather than explicitly. This prevents the user reading the book from being excessively distracted by the scene image JG, thereby providing the user with a better reading experience without hindering their immersion in reading.

[0031] Furthermore, the content presenter 104 may perform image processing, as necessary, to prevent the reader from excessively drawing attention to the scenery image JG. For example, the content presenter 104 may apply a Gaussian filter to the scenery image JG to reduce the contrast of the image and blur it. This allows the atmosphere of the scene to be conveyed without drawing the reader's attention excessively to the image.

[0032] Furthermore, when overlapping the text information T with the scenery image JG, the content presenter 104 fills the area where the text information T is displayed (the rectangular display area R in FIG. 4 ) with a single color (such as gray) and displays the text in a color (such as black) darker than the display area R so that the text information T is not difficult to read due to the scenery image JG that serves as the background. Furthermore, a process may be performed to make the scenery image JG transparent as the background of the display area R. This allows the text information T to be displayed so that it stands out from the scenery image JG that is drawn as the background, making the text easier to read.

[0033] 5 is a flowchart showing the flow of processing performed by the content presentation system 10 of the embodiment. The text information acquisition unit 101 of the content presentation system 10 acquires text information T (step S10). The scene generation prompt generation unit 102 of the content presentation system 10 generates a scene generation prompt by having a large-scale language model extract scene features from the text information T (step S11). The scene image generation unit 103 of the content presentation system 10 generates a scene image JG by inputting the scene generation prompt to an image generation AI (step S12). The content presentation unit 104 of the content presentation system 10 presents the scene image JG together with the text information (step S13).

[0034] As described above, the content presentation system 10 of the first embodiment includes a text information acquisition unit 101, a scene generation prompt generation unit 102, a scene image generation unit 103, and a content presentation unit 104. The text information acquisition unit 101 acquires text information T. The scene generation prompt generation unit 102 generates a scene generation prompt for drawing a scene of a text corresponding to the text information T from the text information T. The scene image generation unit 103 generates a scene image JG corresponding to the text information T by inputting the scene generation prompt to an image generation model. The content presentation unit 104 presents the scene image JG together with the text information T. In this way, the content presentation system 10 of the first embodiment can present the scene image JG together with the text information T. The scene image JG depicts a scene that does not include details of the content described in the text, such as people or characters that attract the reader's attention, but includes details of which the reader is less likely to notice, such as a landscape. This prevents people and characters in the text from looking different from the image, and their appearances from page to page, which can hinder the reader's immersion in reading. Therefore, images generated by image generation AI can be effectively presented to improve the immersion in reading.

[0035] Furthermore, in the content presentation system 10 of the first embodiment, the scene generation prompt generation unit 102 extracts scene features J related to a scene from text information using a large-scale language model, and generates a scene generation prompt using the extracted scene features J. As a result, in the content presentation system 10 of the first embodiment, the scene features J generated in the large-scale language model can be generated as a scene generation prompt, and the scene generation prompt can be easily generated.

[0036] Furthermore, in the content presentation system 10 of the first embodiment, the scenery image generation unit 103 inputs style information specifying the style of the scenery image JG to the image generation model as a positive prompt PPT2 common to each page included in the book. This allows the content presentation system 10 of the first embodiment to unify the style of the scenery image JG corresponding to the text information T for each page included in the book. Therefore, consistency in the style of the scenery image JG for each page can be achieved, improving the sense of immersion in reading.

[0037] Furthermore, in the content presentation system 10 of the first embodiment, the scenery image generation unit 103 inputs keywords that are likely to attract the reader's attention, such as phrases that tend to draw attention to details, into the image generation model as negative prompts. This allows the content presentation system 10 of the first embodiment to generate a scenery image JG that does not include elements that tend to draw attention to details, thereby improving the sense of immersion in reading.

[0038] Furthermore, in the content presentation system 10 of the first embodiment, the scenery image generation unit 103 generates the scenery image JG using an image generation model that generates an image without using elements that are likely to attract the reader's attention, such as LECO. This allows the content presentation system 10 of the first embodiment to generate a scenery image JG that does not include elements that tend to draw the reader's attention to details, thereby improving the sense of immersion in reading.

[0039] Furthermore, in the content presentation system 10 of the first embodiment, the content presentation unit 104 presents an image in which the text information T is partially overlapped in the foreground of the scenery image JG. This allows the content presentation system 10 of the first embodiment to display the scenery image JG around the area in which the text information T is displayed, thereby preventing the reader's immersion in reading from being hindered.

[0040] (Embodiment 2) Here, a second embodiment will be described. In this embodiment, the content presentation system 10 is applied to a scene where an electronic book is viewed on a tablet or the like. Note that the scope of the present invention is not limited to the form described in this embodiment, and is merely an example. Furthermore, parts not described in this embodiment are the same as those in the above-mentioned embodiments.

[0041] FIG. 6 is a diagram showing an example in which the content presenter 104 of the second embodiment displays a content image (an image in which a scenery image JG is displayed as the background of text information T) on a tablet terminal.

[0042] The text information acquisition unit 101 acquires, from digitized book data such as an electronic book, text written on each page of the book corresponding to the book data as text information T. In this way, the content presentation system 10 can acquire, for example, text information T corresponding to each page of the electronic book when a user reads using the electronic book. The content presenting unit 104 presents text information T and a scenery image JG on a tablet or the like, for example, a smartphone, a tablet terminal, or a VR (Virtual Reality, virtual space) of a head-mounted display. This makes it possible to apply the content presentation system 10 of this embodiment when reading using a conventional electronic book system, thereby providing a reading experience that does not impede the sense of immersion. For example, the content presentation system 10 may be linked to a conventional electronic book system so that each time a user turns a page of an electronic book, the electronic book system notifies the content presentation system 10 of the operation signal. As a result, the content presentation system 10 knows the page the user is reading and generates a scenery image JG corresponding to the text written on that page. The content presentation system 10 outputs the generated scenery image JG to the electronic book system, and the text information T is displayed together with the scenery image JG in the display column of the electronic book. Alternatively, the content presentation system 10 downloads the book data all at once, and uses the downloaded book data to generate a scene image JG corresponding to each page, so that each time the user turns a page of the electronic book, the scene image JG corresponding to that page is displayed together with the text information T.

[0043] As described above, in the content presentation system 10 of the second embodiment, the text information acquisition unit 101 acquires the text information T shown on each page of the electronic book using the electronic data of the electronic book. As a result, in the content presentation system 10 of the embodiment, when the user uses the electronic book, it becomes possible to present a scene image JG that matches the content written on each page without impairing the immersive feeling of reading.

[0044] Furthermore, in the content presentation system 10 of the second embodiment, the content presentation unit 104 displays the scenery image JG and the text information T on at least one of a smartphone, a tablet terminal, and a head-mounted display. As a result, the content presentation system 10 of the embodiment can display the scenery image JG and the text information T on any display device that the user uses to read an e-book, thereby improving convenience.

[0045] (Embodiment 3) Here, a third embodiment will be described. In this embodiment, the content presentation system 10 is applied to reading using an analog physical book (hereinafter referred to as an analog book BK). Note that the scope of the present invention is not limited to the form described in this embodiment, and is merely an example. Furthermore, parts not described in this embodiment are the same as those in the above-mentioned embodiments.

[0046] 7 is a diagram showing an example of an image projected by the content presentation unit 104 of the third embodiment. In this diagram, the content presentation unit 104 projects a scene image JG into an appreciation space using a projector PJ. The appreciation space may be, for example, a screen SC such as a 360-degree dome-shaped theater. A camera C is installed above the 360-degree dome-shaped screen, and captures an image of the double-page spread of an analog book being read by a user. The camera C captures a double-page spread of an analog book and outputs the captured page image to the text information acquisition unit 101 . The text information acquisition unit 101 recognizes characters that make up the text captured in the page image by OCR (Optical Character Recognition / Reader) and acquires the recognition result as text information T. Alternatively, the text information acquisition unit 101 may perform character recognition on an area in the page image containing characters indicating the title of the book and numbers indicating the page numbers of the facing pages, and acquire the book title and page numbers based on the character recognition results.The text information acquisition unit 101 may acquire text information T corresponding to the text written on the page by, for example, accessing a database in which book data is stored, based on the acquired book title and page number.

[0047] As described above, in the third embodiment, the text information acquisition unit 101 acquires text information T from page images of each page captured while reading the analog book BK. For example, the text information acquisition unit 101 acquires text information T for each page by character recognition of the characters of the text written on the page and estimating which page is open from the page image captured by a camera (for example, camera C in FIG. 6) installed in a position where the characters can be recognized. As a result, in the third embodiment, it is possible to present a scene image JG that matches the content written on each page while reading the analog book BK without losing the sense of immersion in reading.

[0048] Furthermore, in the content presentation system 10 of the third embodiment, the content presentation unit 104 presents the scene image JG by projecting the scene image JG into the appreciation space using a projector. As a result, in the content presentation system 10 of the embodiment, while the user is reading an analog book BK in the appreciation space, the scene image JG can be projected over the entire appreciation space every time the user turns a page, providing a highly immersive reading experience that makes the user feel as if they are entering the world of the book.

[0049] All or part of the content presentation system 10 in the above-described embodiment may be implemented by a computer. In this case, a program for implementing the functions may be recorded on a computer-readable recording medium, and the program may be loaded into a computer system and executed. Note that the term "computer system" as used herein includes hardware such as an operating system and peripheral devices. Furthermore, the term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into a computer system. Furthermore, the term "computer-readable recording medium" may also include devices that dynamically store programs for a short period of time, such as communication lines used when transmitting programs over a network such as the Internet or a telephone line, or devices that store programs for a fixed period of time, such as volatile memory within a computer system that serves as a server or client. The program may be designed to implement some of the functions described above, or may be capable of implementing the functions in combination with a program already stored in the computer system, or may be implemented using a programmable logic device such as an FPGA.

[0050] Although an embodiment of the present invention has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of the present invention. [Explanation of symbols]

[0051] 10...Content presentation system 101…Text information acquisition unit 102...scene generation prompt generation unit 103...Scenery image generation unit 104...Content presentation section

Claims

1. a text information acquisition unit that acquires text information; a scene generation prompt generating unit that generates, from the text information, a scene generation prompt for drawing a scene of a text corresponding to the text information; a scene image generation unit that generates a scene image, which is a still image or a moving image, corresponding to the text information by inputting the scene generation prompt into an image generation model; a content presenting unit that presents the scenery image together with the text information; A content presentation system comprising:

2. the scene generation prompt generation unit extracts scene features related to a scene from the sentence information using a large-scale language model, and generates the scene generation prompt using the extracted scene features. The content presentation system according to claim 1 .

3. the scenery image generation unit inputs style information specifying a style of the scenery image into an image generation model as a positive prompt common to each page included in the book; The content presentation system according to claim 1 .

4. The scene image generation unit inputs keywords that attract a large degree of reader attention into an image generation model as negative prompts. The content presentation system according to claim 1 .

5. the scene image generation unit generates the scene image using an image generation model that generates an image without using elements that attract a large degree of reader attention; The content presentation system according to claim 1 .

6. the text information acquisition unit acquires the text information shown on each page of the electronic book using electronic data of the electronic book; The content presentation system according to claim 1 .

7. the content presenting unit presents an image in which the text information is partially overlapped in the foreground of the scenery image; The content presentation system according to claim 1 .

8. the content presenting unit presents the text information and the scenery image on at least one of a smartphone, a tablet terminal, and a head-mounted display; The content presentation system according to claim 1 .

9. the text information acquisition unit acquires the text information from a page image obtained by capturing an image of a page in a book; The content presentation system according to claim 1 .

10. the content presentation unit presents the scenery image by projecting it into an appreciation space using a projector. The content presentation system according to claim 1 .

11. A content presentation method performed by a computer, comprising: A text information acquisition unit acquires text information, a scene generation prompt generation unit generates, from the text information, a scene generation prompt for drawing a scene of a sentence corresponding to the text information; a scene image generation unit inputting the scene generation prompt into an image generation model to generate a scene image, which is a still image or a moving image, corresponding to the text information; a content presenting unit presenting the scenery image together with the text information; How content is presented.

12. On the computer, Get text information, generating a scene generation prompt from the text information to draw a text scene corresponding to the text information; generating a scene image, which is a still image or a moving image, corresponding to the text information by inputting the scene generation prompt into an image generation model; presenting the scenery image together with the text information; program.

Images