Image generation device, image generation method, and image generation program

Abstract

This invention provides an image generation device, an image generation method, and an image generation program that enable users to easily and automatically generate catchphrases and banner images. [Solution] The system includes an acquisition unit that acquires input information describing the banner image to be generated; a catchphrase generation unit that generates a catchphrase using a large-scale language model based on the input information; an image generation unit that creates an image generation prompt using a large-scale language model based on the input information and generates a background image using an image generation model based on the input information and the image generation prompt; a banner generation unit that generates a banner image from the catchphrase generated by the catchphrase generation unit and the background image generated by the image generation unit; and a display unit.

Description

Technical Field

[0001] The present disclosure relates to an image generation device, an image generation method, and an image generation program, and particularly to an image generation device, an image generation method, and an image generation program for generating a banner image.

Background Art

[0002] In recent years, devices for automatically generating catchphrases and images for advertising banners have been increasing. Patent Document 1 discloses a system for generating a promotional text according to a user's input. Patent Document 2 discloses a device for generating an image according to a user's input.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the system disclosed in Patent Document 1 only generates a promotional text, and does not consider the arrangement of multiple promotional texts within an advertising banner or the arrangement, etc. Also, according to the device disclosed in Patent Document 2, it is not possible to generate an image that sufficiently meets the purpose only from the text indicating the resulting image as the user's input for generating the image.

[0005] In view of the above problems, an object of the present invention is to provide an image generation device, an image generation method, and an image generation program that can easily automatically generate a catch copy and a banner image by a user.

Means for Solving the Problems

[0006] A first aspect of the present invention is an image generation apparatus comprising: an acquisition unit that acquires input information describing a banner image to be generated; a catchphrase generation unit that generates a catchphrase using a large-scale language model based on the input information; an image generation unit that creates an image generation prompt using a large-scale language model based on the input information and generates a background image using an image generation model based on the input information and the image generation prompt; and a banner generation unit that generates a banner image from the catchphrase generated by the catchphrase generation unit and the background image generated by the image generation unit.

[0007] In a first embodiment of the present invention, the input information may be either a text description of an advertising banner image desired by the user, or the address of a landing page.

[0008] In a first embodiment of the present invention, the catchphrase generation unit may generate a main catchphrase and a sub-catchphrase as catchphrases.

[0009] In a first embodiment of the present invention, the image generation prompt may be created by an image generation unit that searches for existing images used on the web using a web search engine based on input information, and then uses an image text generation model to reference the existing images obtained from the search.

[0010] In a first embodiment of the present invention, the system may further include a user editing unit that accepts user specifications for editing a banner image.

[0011] In a first embodiment of the present invention, the catchphrase is output as a text box, and the specified editing may include editing the text box.

[0012] In a first embodiment of the present invention, the catchphrase generation unit generates multiple catchphrases and displays the generated multiple catchphrases on the display unit, the input unit receives input of a catchphrase to be selected from the multiple generated catchphrases by the user, and the banner image may be generated using the selected catchphrase received by the input unit.

[0013] In a first embodiment of the present invention, the image generation unit generates a plurality of background images and displays the generated plurality of background images on the display unit, the input unit receives input of a background image selected by the user from the plurality of generated background images, and the banner image may be generated using the selected background image received by the input unit.

[0014] A second aspect of the present invention is an image generation method comprising: an acquisition step of acquiring input information describing a banner image to be generated; a catchphrase generation step of generating a catchphrase using a large-scale language model based on the input information; an image generation step of creating an image generation prompt using a large-scale language model based on the input information and generating a background image using an image generation model based on the input information and the image generation prompt; and a banner generation step of generating a banner image from the catchphrase generated in the catchphrase generation step and the background image generated in the image generation step.

[0015] A third aspect of the present invention is an image generation program that enables a computer to implement an acquisition function for acquiring input information describing a banner image to be generated; a catchphrase generation function that generates a catchphrase using a large-scale language model based on the input information; an image generation function that creates an image generation prompt using a large-scale language model based on the input information and generates a background image using an image generation model based on the input information and the image generation prompt; and a banner generation function that generates a banner image from the catchphrase generated in the catchphrase generation function and the background image generated in the image function.

[0016] According to the present invention, it is possible to provide an image generation device, an image generation method, and an image generation program that can easily generate catchphrases and banner images by a user.

Brief Description of Drawings

[0017] [Figure 1] It is a block diagram showing an example of the configuration and functional units of the image generation device according to this embodiment. [Figure 2] It is a schematic diagram showing an existing advertisement image when the image generation unit generates an image. [Figure 3] It is the result of textifying the elements of the obtained advertisement image shown in FIG. 2. [Figure 4] It is a schematic diagram of an example of a template used by the banner generation unit. [Figure 5] It is a diagram showing how the input information is displayed on the display unit. [Figure 6] It is a flowchart for explaining the image generation method according to this embodiment. [Figure 7] It is a diagram showing that a plurality of background image candidates generated by the image generation unit are displayed on the display unit. [Figure 8] It is an example of a banner image generated by the image generation device according to this embodiment. [Figure 9] It is another example of a banner image generated by the image generation device according to this embodiment. [Figure 10] It is another example of a banner image generated by the image generation device according to this embodiment. [Figure 11] It is a flowchart for explaining the image generation method according to this embodiment.

Mode for Carrying Out the Invention

[0018] Next, embodiments of the present invention will be described with reference to the drawings. In the drawings of the embodiments, identical or similar parts are denoted by the same or similar reference numerals. However, it should be noted that the drawings are schematic and the relationships with planar dimensions etc. may differ from those in reality. Therefore, specific dimensions should be determined by referring to the following explanation. Furthermore, it goes without saying that there are parts in the drawings where the relationships and ratios of dimensions differ from those of other parts.

[0019] Furthermore, the embodiments are illustrative of apparatus and methods for realizing the technical concept of the present invention, and the technical concept of the present invention does not limit the configuration, arrangement, layout, etc., of each component to those described below. The technical concept of the present invention can be modified in various ways within the technical scope defined by the claims described in the patent claims.

[0020] (First embodiment) The image generation device according to this embodiment can automatically generate a catchphrase and a banner image in response to user input. Furthermore, the image generation device according to this embodiment allows the user to freely edit the position and content of the catchphrase and background image of the generated banner image.

[0021] Figure 1 shows an example of the configuration and functions of the image generation device 10 according to this embodiment. The image generation device 10 shown in Figure 1 includes a CPU 21 for executing various calculations, a ROM 22 for storing processing programs, a RAM 23 for storing data, a storage unit 24 for storing various data and calculation results, an I / O (input / output interface) 25, a display unit 26, an input unit 27, and the like.

[0022] I / O25 is an interface, buffer, etc., for communication (transmitting and receiving).

[0023] The image generation device 10 is a variety of electronic computer (computational resource), such as a mobile terminal, personal computer (PC), mainframe, workstation, or cloud computing system. In the example shown in Figure 1, the display unit 26 is a display, and the input unit 27 is a keyboard, mouse, etc. for input, which accepts user input.

[0024] Furthermore, the block diagram in Figure 1 shows the functional units within the CPU 21. When each functional unit of the CPU 21 is implemented by software, the CPU 21 implements these functions by executing instructions from the program, which is the software that realizes each function. Specifically, it includes an acquisition unit 210, a catchphrase generation unit 211, an image generation unit 212, a banner generation unit 213, an image editing unit 214, and so on.

[0025] The acquisition unit 210 acquires information describing the banner image to be generated by the image generation device 10, as requested by the user, via the I / O 23 or the input unit 27. The information describing the banner image to be generated by the image generation device 10 is used as input information.

[0026] The input information is, for example, text, or information relating to a landing page that can be displayed from a banner image generated by the image generation device 10. If the input information is text, the text describes the banner image desired by the user. The text may also include, for example, the purpose of the advertisement, the target users, characteristics, etc. If the input information is information relating to a landing page, the information relating to the landing page is, for example, the address of the web page of the landing page. The following describes the case where the input information is text, and the case where the input information is information relating to a landing page will be described in the second embodiment.

[0027] The catchphrase generation unit 208 generates a catchphrase in text format using a general-purpose large-scale language model based on the input information acquired by the acquisition unit 210. The catchphrase generation unit 208 may also generate a single catchphrase as the first catchphrase.

[0028] The catchphrase generation unit 208 may generate multiple catchphrases. Typically, the catchphrases in advertising banners consist of multiple texts. Furthermore, generally, the multiple catchphrases in an advertising banner consist of a main catchphrase and sub-catchphrases. Additionally, in advertising banners, the main catchphrase is displayed in the largest text size, while the sub-catchphrases are displayed in smaller text sizes compared to the main catchphrase. The catchphrase generation unit 208 generates multiple catchphrases, and we will refer to the main catchphrase as the first catchphrase, the sub-catchphrases as the second catchphrase, the third catchphrase, and so on, in order of importance. In the following explanation, we will assume that the catchphrase generation unit 208 generates multiple catchphrases.

[0029] The catchphrase generation unit 208 may generate multiple first catchphrases, second catchphrases, third catchphrases, ... The catchphrase generation unit 208 may display each of the generated multiple first catchphrases, multiple second catchphrases, multiple third catchphrases, ... on the display unit 26 as the nth catchphrase candidate (where n is an integer of 1 or more), and the user may select the nth catchphrase from among the multiple nth catchphrase candidates displayed on the display unit 26 to be used by the banner generation unit 213 to generate a banner. In this case, the input unit 27 may accept input from the user to select the nth catchphrase from among the multiple nth catchphrase candidates.

[0030] The catchphrase generation unit 208 may change the importance n of each of the multiple catchphrases it has generated, that is, it may change the nth catchphrase to the mth catchphrase (n ≠ m). The catchphrase generation unit 208 may be configured to allow the user to select a desired catchphrase from the nth catchphrases selected for use in generating a banner by the banner generation unit 213, and to change its importance n. In this case, the input unit 27 may accept input from the user to select a desired catchphrase from the nth catchphrases and change its importance n.

[0031] The image generation unit 212 generates a background image for the banner image using a general-purpose text image generation model that generates images from text, based on the input information acquired by the acquisition unit 210.

[0032] When the image generation unit 212 generates a background image, it cannot provide sufficiently specific instructions to generate a background image that matches the user's desired background image based solely on the user's input information. In order to generate a background image that the user desires, information that specifically describes the background image the user desires is necessary. For example, the background image desired by the user may vary depending on the purpose of use, such as whether it is for advertising, for use in public settings such as corporate web pages, or for use in private blogs. However, specific instructions for generating a background image that conforms to the purpose of use are insufficient based solely on the input information. In order to improve the relationship between the background image desired by the user and the background image generated by the image generation unit 212, the image generation unit 212 generates the image using the following procedure.

[0033] The image generation unit 212 creates an image generation prompt using a large-scale language model based on the input information acquired by the acquisition unit 210. Next, based on the input information acquired by the acquisition unit 210 and the created image generation prompt, it generates a background image using a general-purpose image generation model that generates images from text.

[0034] The image generation unit 212 may generate a background image using a general-purpose image generation model that generates images from text, based on the input information acquired by the acquisition unit 210, the created image generation prompt, and, when the catchphrase generation unit 208 generates multiple catchphrases such as the first catchphrase, second catchphrase, third catchphrase, ..., the catchphrase selected by the user from among the generated catchphrases.

[0035] In order to improve the relationship between the input information acquired by the acquisition unit 210 and the background image generated by the image generation unit 212, the image satellite device 10 according to this embodiment may generate the background image in the following ways. Based on the input information acquired by the acquisition unit 210, the image generation unit 212 may first search for existing images used on the web using an existing web search engine. Next, the image generation unit 212 may refer to the image obtained through the search and use a general-purpose image-text generation model to decompose the elements of the image, and convert the decomposed elements into text. The image generation unit 212 may use the obtained elements as an image generation prompt and generate the background image using a general-purpose text-image generation model. The image generation unit 212 may also generate the background image using a general-purpose image generation model that generates images from text, based on the input information acquired by the acquisition unit 210 and the obtained elements.

[0036] Figure 2 shows an example where, when a user requests an advertising banner for a mobile device, the image generation unit 212 searches the web using an existing search engine based on the input information acquired by the acquisition unit 210 and obtains an existing advertising image used on the web. Figure 3 shows the elements of the obtained advertising image converted into text. Figure 2 shows a photo of a woman talking on a mobile device, along with a catchphrase, etc. As shown in Figure 3, "Logo: The word "Mobile" is written in white in the upper right corner of the image. This is the company's logo, contrasting with the pink background.", "Main message: The words "Don't worry about your wallet" are displayed in large white letters at the top center. Just below that, the words "I want to use the data!" are written in slightly smaller white letters.", ... are the text versions of each element of the advertising image shown in Figure 2. Based on the input information acquired by the acquisition unit 210 and the text shown in Figure 3, the image generation unit 212 generates a banner image using a general-purpose image generation model that generates images from text.

[0037] The image generation unit 212 may generate multiple background images based on the input information acquired by the acquisition unit 210. The image generation unit 212 may display the generated multiple background images as background image candidates on the display unit 26, and the user may select a background image from among the multiple background image candidates displayed on the display unit 26 to be used by the banner generation unit 213 to generate a banner image. In this case, the input unit 27 may accept input from the user to select a background image from among the background candidates.

[0038] The banner generation unit 213 generates a banner image from the catchphrase text generated by the catchphrase generation unit 208 and the background image generated by the image generation unit 212.

[0039] The banner generation unit 213 may display a CTA (Call To Action) within the banner image generated by the image generation unit 212 when generating the banner image. A CTA refers to an element on the banner screen that guides visitors to the site, and on the banner image, it is displayed in the form of a button, link, etc. For example, clicking a button-shaped CTA on the banner screen that says "Request materials here" will display a web page for requesting materials. The input unit 27 may accept input from the user regarding information about the CTA to be displayed. This information about the CTA to be displayed includes the URL of the link destination, the text to be displayed on the CTA such as "Request materials here", the shape of the button, etc.

[0040] When generating a banner image, the banner generation unit 213 may display a logo within the banner image generated by the image generation unit 212. The logo may be acquired, for example, by the acquisition unit 210.

[0041] When the banner generation unit 213 generates a banner image using the catchphrase text generated by the catchphrase generation unit 208 and the background image generated by the image generation unit 212, it generates the banner based on a pre-specified banner design, i.e., a design template. The design template is a design in which the position of the text boxes relative to the background image, the text color, the background color of the text boxes, etc., are predetermined. Based on the design template, the banner generation unit 213 can generate a banner using the catchphrase and the banner image.

[0042] Figure 4 shows a schematic diagram of an example of a template used by the banner generation unit 213. As shown in Figure 4, the placement of the background image 46, subject information 42, first catchphrase 43, second catchphrase 44, and CTA 45 is predetermined in template 41. The subject information 42 displays the subject of the advertising banner. The subject information of the advertising banner may be, for example, the logo of the advertiser company.

[0043] The image generation device 10 according to this embodiment may store a plurality of design templates in the storage unit 24, and the banner generation unit 213 may generate a plurality of banners based on the plurality of design templates stored in the storage unit 24. The banner generation unit 213 displays the plurality of generated banners as a plurality of banner candidates on the display unit 26, and a banner may be selected by the user from among the banner candidates.

[0044] The banner generation unit 213 places the catchphrase text generated by the catchphrase generation unit 208 at a predetermined position within the background image generated by the image generation unit 212, based on a design template. The image generation device 10 according to this embodiment may be configured to allow the user to edit the catchphrase text of the banner image generated by the banner generation unit 213. For example, the image generation device 10 may further include an image editing unit 214 that accepts user requests for editing of the banner image.

[0045] The catchphrase generated by the catchphrase generation unit 208 may be output as a text box, and the image editing unit 214 may allow the user to edit the text box based on their input. Here, editing of the text box may allow editing of at least one of the following: the placement of the catchphrase text box relative to the background image, the background of the text box, the font color, type, font size, bolding, line spacing, and character spacing of the text box. In this case, the input unit 27 may accept the user's specifications for editing the banner and transmit the received content to the image editing unit 214.

[0046] When the banner generation unit 213 generates a banner image based on a design template, the catchphrase text is not necessarily positioned on the background image. For example, the catchphrase and background image of the banner image generated by the banner generation unit 213 may not overlap, or may only partially overlap. An example of a banner image generated by the banner generation unit 213 where the catchphrase and background image do not overlap will be described later with reference to Figure 9.

[0047] Next, the operation of the image generation device 10 according to this embodiment will be explained with reference to Figures 5 to 8, using specific examples.

[0048] First, the acquisition unit 210 receives information from the user via I / O 23 or input unit 27 that describes the banner to be generated by the image generation device 10, as desired by the user. For example, if input unit 27 is an input keyboard, the user uses the keyboard to input information describing the banner in bullet points.

[0049] Figure 5 shows how the input information describing the banner, entered by the user, is displayed on the display unit 25. For example, the user might enter: "Purpose: An advertising banner with a conversion (CV) of requesting information about an operational management app for paid sports school management organizations," "Target users: Operators of paid club activities or sports schools," "Features: A monthly fee is determined for each member, and everything from communication with parents to payment and management of monthly fees can be centralized," "Keywords related to this banner: Monthly fee of 100 yen per member, one year free for those who apply by December."

[0050] In the example described above, the information describing the banner to be generated by the image generation device 10 was entered in bullet points, including the purpose of the banner, the target users of the banner, the characteristics of the banner, and keywords related to the banner, but it is not limited to this. For example, the information describing the banner may be entered in sentences rather than bullet points, or keywords may be listed and entered.

[0051] When the acquisition unit 210 receives information describing the banner, it transmits the received information to the catchphrase generation unit 208. Based on the input information acquired by the acquisition unit 210, the catchphrase generation unit 208 generates a catchphrase in text format using a general-purpose large-scale language model.

[0052] Figure 6 shows how the candidates for the first catchphrase and the second catchphrase generated by the catchphrase generation unit 208 are displayed on the display unit 26. The catchphrase generation unit 208 generates multiple first catchphrases and multiple second catchphrases, and displays them on the display unit 26 as candidates 61 for the first catchphrase and candidates 62 for the second catchphrase. The input unit 27 accepts input from the user to select the first catchphrase 63 from the multiple candidates 61 for the first catchphrase and the second catchphrase 64 from the multiple candidates 62 for the second catchphrase. In the example shown in Figure 5, the user has selected "No more worries about member management" as the first catchphrase 63 and "Reduce the burden of running clubs and schools!" as the second catchphrase 64.

[0053] As shown in the example in Figure 6, all of the first catchphrase candidates 61 consist of 20 characters or less, and all of the second catchphrase candidates 62 consist of 40 characters or less. Typically, the main catchphrase is short, around 20 characters or less, and contains content that will attract the user's interest. The sub-catchphrase, on the other hand, consists of more characters than the main catchphrase, around 40 characters or less, and concisely explains the content of the banner.

[0054] The multiple first catchphrase candidates 61 and the multiple second catchphrase candidates 62 shown in Figure 6 have character limits of 20 characters and 40 characters respectively, and the catchphrases generated by the catchphrase generation unit 208 are limited in character count. However, this character limit may be changed by the user.

[0055] Next, when the acquisition unit 210 receives information describing the banner, it transmits the received information to the image generation unit 212. Based on the input information acquired by the acquisition unit 210, the image generation unit 212 generates a banner image using a general-purpose text-to-image generation model that generates images from text. In the example shown in Figure 7, the image generation unit 212 generates multiple candidate banner images.

[0056] Figure 7 shows how multiple background image candidates 71a to 71d generated by the image generation unit 212 are displayed on the display unit 26. Multiple background image candidates 71a to 71d generated by the image generation unit 212 are displayed on the display unit 26, and background image candidate 71b is selected by the user.

[0057] The banner generation unit 213 generates a banner image using the first catchphrase 63 and the second catchphrase 64 selected in the example shown in Figure 6, and the candidate background image 71b selected in the example shown in Figure 7 as the background image. An example of a banner image generated by the banner generation unit 213 is shown in Figure 8. The banner image 81 shown in Figure 8 has a background image 82 using the candidate background image 71b selected in the example shown in Figure 7, with the first catchphrase 63 and the second catchphrase 64 selected in the example shown in Figure 6, and a CTA 65 placed on top. Information regarding the CTA 65 is assumed to have been input in advance by the user via the input unit 27.

[0058] The banner image 81 shown in Figure 8 has the text boxes 83 and 84 for the first catchphrase 63 and second catchphrase 64 placed directly on top of the background image 82, which looks unnatural. The line breaks within the text boxes 83 and 84 for the first catchphrase 63 and second catchphrase 64 are not specified, making it difficult for the reader to read.

[0059] Here, the image editing unit 214 can eliminate any unnatural appearance by making the backgrounds of text boxes 83 and 84 transparent based on user input. Furthermore, the image editing unit 214 can edit the placement of text boxes 83 and 84 on the background image 82, as well as the font color, font size, and line break position of the first catchphrase 63 and the second catchphrase 64, based on user input.

[0060] Figure 9 shows an example of a banner image generated by the banner generation unit 213 using a template. The banner image 90 shown in Figure 9 has a background image 91 which is a photograph of a part of a living room, and displays a button with a link to a webpage where you can request information from, for example, a renovation company, with the first catchphrase 92 being "Fully renovate the interior of your home," the second catchphrase 93 being "Make your daily life more luxurious," and the CTA 94 being "Click here for details." The sponsor information is not displayed in Figure 9. The banner shown in Figure 9 was generated using a template in which the position of the text boxes, the text color, the background color of the text boxes, and the position of the background image are predetermined, and the text boxes have an editing function.

[0061] Figure 10 shows another example of a banner image generated by the banner generation unit 213 using a template. The banner image 100 shown in Figure 10 has a background image 101 which is a photograph of a children's soccer team playing a soccer game, a first catchphrase 102 which reads "Easy payment processing", a second catchphrase 103 which reads "For those struggling with sports club management", a CTA 104 which reads "Request information", a button with a link to a webpage for requesting information from a company that provides an application for managing school operations, for example, and the company name 105 which displays "SApr", the logo of the application for managing school operations. The banner image shown in Figure 10 is generated using a template in which the background image 101, first catchphrase 102, second catchphrase 103, CTA 104, and company name 105 are separated so that they do not overlap with each other.

[0062] The image generation method according to this embodiment will be explained with reference to the flowchart in Figure 11.

[0063] In step S1101, the acquisition unit 210 acquires information as input information that describes the banner image to be generated by the image generation device 10, as requested by the user (acquisition step).

[0064] In step S1102, the catchphrase generation unit generates a catchphrase using a large-scale language model based on the input information (catchphrase generation step).

[0065] In step S1103, the image generation unit creates an image generation prompt using a large-scale language model based on the input information. (Image generation prompt creation step).

[0066] In step S1104, the image generation unit 212 generates a background image using a general-purpose image generation model that generates images from text, based on the input information acquired by the acquisition unit 210 and the created image generation prompt. (Background image generation step).

[0067] In step S1105, the banner generation unit generates a banner image from the generated background image and the generated catchphrase (banner image generation step).

[0068] In step S1106, the display unit 26 displays the banner image generated by the banner generation unit (display step).

[0069] (Second embodiment) In the first embodiment, the input information acquired by the acquisition unit 210 was a text document. In the second embodiment, the case where the input information acquired by the acquisition unit 210 is a URL of a landing page is described. In this case, the advertising banner generated by the image generation device according to the second embodiment displays a CTA that specifies the landing page as the link destination.

[0070] The image generation apparatus according to the second embodiment further includes an image text generation unit compared to the image generation apparatus 10 according to the first embodiment.

[0071] The catchphrase generation unit 208 generates a catchphrase in text format using a general-purpose large-scale language model, based on at least one of the landing page's meta information, meta title, and meta description information, instead of the input information acquired by the acquisition unit 210. Here, meta information is HTML code that conveys information about the website to search engines, browsers, etc., the meta title is the title of the website displayed in search engine results, and the meta description is information that describes the overview of the website.

[0072] Similar to the catchphrase generation unit 208, the image generation unit 212 generates a banner image using a general-purpose text-to-image generation model that generates images from text, based on at least one of the landing page's meta information, meta title, and meta description information, instead of the input information acquired by the acquisition unit 210.

[0073] The banner generation unit 213 generates a banner from the catchphrase text generated by the catchphrase generation unit 208 and the banner image generated by the image generation unit 212, similar to the image generation device 10 according to the first embodiment.

[0074] (Third embodiment) Images used on web pages of companies, hospitals, etc., are typically chosen to match the company, facility, and content of the web page. However, when articles are frequently posted on the web page, different images may be needed each time, or a large number of similar images may be required. The image generation device according to the third embodiment generates similar images from existing images.

[0075] The image generation device according to the third embodiment may have the same configuration as the image generation device 10 according to the first embodiment, except for the functional parts within the CPU 21. The image generation device according to the fourth embodiment includes an acquisition unit 210 and an image generation unit 212 among the functional parts within the CPU 21 of the image generation device 10 according to the first embodiment.

[0076] The acquisition unit 210 acquires the banner image via I / O 23 or input unit 27.

[0077] The image generation unit 212 references the banner image acquired by the acquisition unit 210 and uses a general-purpose image-text generation model to decompose the elements of the banner image and convert the decomposed elements into text. Based on the input information acquired by the acquisition unit 210 and the obtained elements, the image generation unit 212 generates a banner image using a general-purpose text-image generation model that generates images from text.

[0078] As stated above, the present invention naturally includes various embodiments and the like that are not described herein. Therefore, the technical scope of the present invention is determined solely by the inventive features relating to the claims that are reasonable based on the above description. [Explanation of Symbols]

[0079] 10 Image generation device 21 CPU 22 ROM 23 RAM 24 Memory section 25 I / O 26 Display section 27 Input section 210 Acquisition Department 211 Catchphrase Generation Department 212 Image Generation Unit 213 Banner Generation Section 214 Image Editing Department 41 Templates 42, 95 Subject designation 43, 63, 92, 102 First catchphrase 44, 64, 93, 103 Second catchphrase 45, 65, 94, 104 CTA 46, 82, 91, 101 Background images 61. First Catchphrase Candidate 62. Second Catchphrase Candidate 71a~71d Candidate background images 100 banner images 105 Subject designation

Claims

1. An acquisition unit that obtains input information describing the banner image to be generated, A catchphrase generation unit generates a catchphrase using a large-scale language model based on the aforementioned input information, Based on the aforementioned input information, an image generation prompt is created using a large-scale language model. An image generation unit generates a background image using an image generation model based on the input information and the image generation prompt, A banner generation unit generates a banner image from the catchphrase generated by the catchphrase generation unit and the background image generated by the image generation unit. A display unit that displays the generated banner image and An image generation device characterized by comprising the following features.

2. The image generating apparatus according to claim 1, characterized in that the input information is either a text description of an advertising banner image desired by the user, or the address of a landing page.

3. The image generation apparatus according to claim 1, characterized in that the catchphrase generation unit generates a main catchphrase and a sub-catchphrase as the catchphrase.

4. The image generation apparatus according to claim 1, characterized in that the image generation unit searches for existing images used on the web using a web search engine based on the input information, and creates the image generation prompt using an image text generation model by referring to the existing images obtained through the search.

5. The image generation apparatus according to claim 1, further comprising a user editing unit that accepts user specifications for editing the banner image.

6. The image generation apparatus according to claim 5, characterized in that the catchphrase is output as a text box, and the specified editing includes editing the text box.

7. The image generation apparatus according to claim 1, further comprising an input unit, wherein the catchphrase generation unit generates a plurality of catchphrases and displays the generated plurality of catchphrases on the display unit, the input unit receives input of a catchphrase to be selected from the plurality of generated catchphrases by the user, and the banner image is generated using the selected catchphrase received by the input unit.

8. The image generation apparatus according to claim 1, further comprising an input unit, wherein the image generation unit generates a plurality of background images and displays the plurality of generated background images on the display unit, the input unit receives input of a background image selected by the user from the plurality of generated background images, and the banner image is generated using the selected background image received by the input unit.

9. In an image generation device, An acquisition step to obtain input information that describes the banner image to be generated, A catchphrase generation step that generates a catchphrase using a large-scale language model based on the aforementioned input information, Based on the aforementioned input information, an image generation prompt is created using a large-scale language model. An image generation unit generates a background image using an image generation model based on the input information and the image generation prompt, A banner generation step that generates a banner image from the catchphrase generated in the catchphrase generation step and the background image generated in the image generation step, A display step of displaying the generated banner image and An image generation method characterized by comprising the following features.

10. On the computer, A function to obtain input information that describes the banner image to be generated, A catchphrase generation function that generates catchphrases using a large-scale language model based on the aforementioned input information, Based on the aforementioned input information, an image generation prompt is created using a large-scale language model. An image generation unit generates a background image using an image generation model based on the input information and the image generation prompt, A banner generation function generates a banner image from the catchphrase generated in the catchphrase generation function and the background image generated in the image generation function. A display function that displays the generated banner image and An image generation program that achieves this.

Images