Game card generation system

Abstract

This invention provides a game card generation system that can improve the consistency between character image information and character details when generating character game card information (character image information and character details) from text information related to game cards. [Solution] When arbitrary text information related to a game card is input to the game card generation system 1, it generates image information of a character using a first learning model for image generation, and generates detailed information about the character from the image information using a second learning model different from the first learning model. Then, it combines the image information of the character and the detailed information about the character to generate game card information for the character.

Description

[Technical Field]

[0001] The present invention relates to a game card generation system that generates character game card information (character image information and detailed character information) from text information related to game cards. [Background Art]

[0002] In recent years, learning models have been used in various situations. For example, an apparatus that uses a machine learning model when generating an active electronic card provided with a mechanism (interaction mechanism) that directly interacts with a receiver who receives card information has been proposed (see, for example, Patent Document 1). [Prior Art Documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent No. 7093585 [Summary of the Invention] [Problem to be Solved by the Invention]

[0004] However, conventionally, no system has been proposed that generates character game card information (character image information and detailed character information) from text information related to game cards.

[0005] Therefore, one might consider simply using a learning model to generate character game card information (character image information and character details) from text information about game cards. However, when generating character game card information (character image information and character details) from text information about game cards using only one learning model, there is a problem in that even for the same character, the character image information and character details may not match and may be scattered, resulting in low consistency between the character image information and character details.

[0006] The present invention has been made in view of the above problems, and aims to provide a game card generation system that can improve the consistency between character image information and character details when generating character game card information (character image information and character details) from text information related to game cards. [Means for solving the problem]

[0007] The game card generation system of the present invention comprises: an input unit into which arbitrary text information relating to a game card is input; a character image estimation unit that uses a first learning model for image generation to estimate and output image information of a character of the game card to be generated, taking the arbitrary text information input to the input unit as input; a character detailed information estimation unit that uses a second learning model different from the first learning model to estimate and output detailed information of the character, taking the image information of the character output from the character image estimation unit as input; and a game card information generation unit that combines the image information of the character output from the character image estimation unit and the detailed information of the character output from the character detailed information estimation unit to generate and output game card information of the character.

[0008] In this configuration, when arbitrary text information about a game card is input, a first learning model (a learning model for image generation) is used to generate image information of the character from the text information, and then a second learning model (a learning model different from the first learning model) is used to generate detailed information about the character from the image information. Then, the game card information of that character is generated by combining the image information of the character generated by the first learning model and the detailed information of the character generated by the second learning model. By generating the game card information of a character (image information of the character and detailed information of the character) using two learning models (the first and second learning models) in this way, it is possible to improve the consistency between the image information of the character and the detailed information of the character compared to simply using a single learning model.

[0009] Furthermore, the game card generation system of the present invention includes a storage unit that stores, in association with, the input text information entered into the input unit and the generated game card information generated by the game card information generation unit based on the input text information, and a text match determination unit that determines whether new text information entered into the input unit matches the input text information stored in the storage unit. The game card information generation unit may output the generated game card information stored in the storage unit in association with the input text information if the text match determination unit determines that the new text information entered into the input unit matches the input text information stored in the storage unit.

[0010] With this configuration, when previously entered text information is entered again, the game card information is not regenerated based on the re-entered text information. Instead, the previously generated game card information generated when that text information was previously entered is output. Depending on the learning model, if game card information is regenerated based on the re-entered text information, the previously generated game card information generated when that text information was previously entered may not be generated, which could lead to a decrease in reproducibility. A decrease in reproducibility can lead to a decrease in user satisfaction. According to the present invention, when previously entered text information is entered again, the previously generated game card information generated when that text information was previously entered is output, thus maintaining reproducibility.

[0011] Furthermore, in the game card generation system of the present invention, in addition to the arbitrary text information, additional information indicating the worldview of the game card to be generated may be input to the input unit, and the character image estimation unit may use the first learning model to estimate and output image information of the character of the game card to be generated, taking the arbitrary text information and the additional information input to the input unit as input.

[0012] With this configuration, by inputting additional information that indicates the worldview of the game card in addition to the text information about the game card, it is possible to generate character image information that reflects the worldview. According to the present invention, detailed information about the character is generated based on the character image information, and as a result, it is possible to generate detailed information about the character that reflects the worldview. In this way, it is possible to generate game card information for a character that reflects the worldview (character image information and character detailed information).

[0013] Furthermore, the game card generation system of the present invention may include an additional learning unit that further trains the first learning model by fine-tuning using image information that has the worldview of the game card to be generated, and the character image estimation unit may use the additionally trained first learning model to estimate and output image information of the character of the game card to be generated, taking the arbitrary text information input to the input unit as input.

[0014] With this configuration, by using image information that embodies the worldview of the game card and further training the first learning model through fine-tuning, it is possible to generate character image information that reflects the worldview. According to the present invention, detailed character information is generated based on the character image information, and as a result, detailed character information that reflects the worldview can be generated. In this way, game card information of a character that reflects the worldview (character image information and character detailed information) can be generated.

[0015] The method of the present invention is a method performed in a game card generation system, the method comprising: an input step in which arbitrary text information relating to a game card is input; a character image output step in which, using a first learning model for image generation, the arbitrary text information input in the input step is used as input to estimate and output image information of a character of the game card to be generated; a character details output step in which, using a second learning model different from the first learning model, the image information of the character output in the character image output step is used as input to estimate and output detailed information of the character; and a game card information generation step in which the image information of the character output in the character image output step and the detailed information of the character output in the character details output step are combined to generate and output game card information of the character.

[0016] This method, as with the system described above, also works as follows: when arbitrary text information about a game card is input, the first learning model (a learning model for image generation) is used to generate image information of the character from the text information, and then the second learning model (a learning model different from the first learning model) is used to generate detailed information about the character from the image information. The image information of the character generated by the first learning model and the detailed information of the character generated by the second learning model are then combined to generate the game card information of that character. By generating the game card information of a character (image information of the character and detailed information of the character) using two learning models (the first and second learning models), it becomes possible to improve the consistency between the image information of the character and the detailed information of the character compared to simply using a single learning model.

[0017] The program of the present invention is a program executed in a game card generation system, the program causing the computer of the game card generation system to execute: an input process in which arbitrary text information relating to a game card is input; a character image output process that uses a first learning model for image generation to estimate and output image information of a character of the game card to be generated, using the arbitrary text information input in the input process as input; a character detailed information output process that uses a second learning model different from the first learning model to estimate and output detailed information of the character, using the image information of the character output in the character image output process as input; and a game card information generation process that combines the image information of the character output in the character image output process and the detailed information of the character output in the character detailed information output process to generate and output game card information of the character.

[0018] According to this program, similarly to the system described above, when arbitrary text information relating to a game card is input, character image information is generated from the text information using a first learning model (a learning model for image generation), and detailed character information is further generated from the character image information using a second learning model (a learning model different from the first learning model). Then, game card information for the character is generated by combining the character image information generated by the first learning model and the detailed character information generated by the second learning model. By generating character game card information (character image information and detailed character information) using two learning models (the first learning model and the second learning model) in this manner, it becomes possible to improve consistency between the character image information and the detailed character information compared to a case where only one learning model is used. [Effects of the Invention]

[0019] According to the present invention, when generating character game card information (character image information and detailed character information) from text information relating to a game card, it is possible to improve consistency between the character image information and the detailed character information. [Brief Description of the Drawings]

[0020] [Figure 1] FIG. 1 is a block diagram showing the configuration of a game card generation system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing an example of estimating character image information. [Figure 3] FIG. 3 is a diagram showing an example of estimating detailed character information. [Figure 4] FIG. 4 is a diagram showing an example of generated game card information. [Figure 5] FIG. 5 is a sequence diagram for explaining an example of the operation of the game card generation system according to an embodiment of the present invention. [Figure 6]It is a block diagram showing the configuration of a game card generation system (modified example). [Figure 7] It is a sequence diagram for explaining the operation of a game card generation system (modified example).

Mode for Carrying Out the Invention

[0021] Hereinafter, a game card generation system according to an embodiment of the present invention will be described with reference to the drawings. In the present embodiment, the case of a game card generation system used for generating competitive game cards and the like is exemplified. The game card generation system of the present embodiment has a function of generating character game card information (character image information and detailed character information) from text information related to game cards. This function is realized by the hardware of the game card generation system cooperating with a program stored in a memory area or the like.

[0022] The configuration of the game card generation system according to the embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the game card generation system of the present embodiment. As shown in FIG. 1, a game card generation system 1 is connected to a user device 2 via a network N such as the Internet. In the present embodiment, the game card generation system 1 is configured by, for example, a cloud server device or the like, and the user device 2 is configured by, for example, a computer device or the like.

[0023] First, the user device 2 will be described. As shown in FIG. 1, the user device 2 includes an input unit 3 such as a keyboard or a mouse, and a display unit 4 such as a display.

[0024] Arbitrary text information related to a game card conceived by a user (for example, "one who destroys everything born from the twilight") is input to the input unit 3 of the user device 2 (see FIG. 2). The text information input from the input unit 3 of the user device 2 is transmitted to the game card generation system 1.

[0025] Furthermore, additional information indicating the worldview of the game card (for example, "characters from a trading card game") may be input to the input unit 3 of the user device 2. The additional information input from the input unit 3 of the user device 2 is transmitted to the game card generation system 1. Note that the additional information is not limited to text information. For example, image information may be input as additional information, or both text information and image information may be input.

[0026] Next, the game card generation system 1 will be described. As shown in Figure 1, the game card generation system 1 includes an input unit 5, an output unit 6, a storage unit 7, and a control unit 8.

[0027] Input unit 5 receives text information transmitted from user device 2, that is, arbitrary text information entered in input unit 3 of user device 2 (for example, "The one born from twilight who destroys all") (see Figure 2). In addition, if additional information (for example, "a character from a trading card game") is entered from input unit 3 of user device 2, the additional information transmitted from user device 2 is entered into input unit 5. Note that arbitrary text information and additional information may also be entered into input unit 5 from sources other than user device 2, such as a desktop PC or notebook PC. For example, arbitrary text information and additional information can be entered into input unit 5 from another user device, such as a smartphone.

[0028] As will be described later, the output unit 6 transmits the game card information of the character generated by the game card generation system 1 to the user device 2.

[0029] Memory unit 7 consists of a large-capacity memory and stores the data and programs necessary to create character game card information from arbitrary text information.

[0030] The control unit 8 is composed of a CPU, MPU, etc., and includes a character image estimation unit 80, a character detailed information estimation unit 81, a game card information generation unit 82, and a text matching determination unit 83 as functional blocks for generating character game card information from arbitrary text information.

[0031] The character image estimation unit 80 has the function of estimating and outputting image information of the character to be generated on the game card, using an image generation learning model (first learning model) as input to the input unit 5, with arbitrary text information input to the input unit 5 as input. Specifically, the character image estimation unit 80 generates a prompt to estimate and output character image information based on the arbitrary text information input to the input unit 5, and inputs the generated prompt to the image generation learning model (first learning model). As a result, the character image estimation unit 80 obtains character image information (character image information estimated from the arbitrary text information) as output from the image generation learning model (first learning model).

[0032] The character image estimation unit 80 may, for example, use an existing large-scale language model such as GPT (Generative Pre-trained Transformer) to generate prompts to be input to the image generation learning model. That is, the character image estimation unit 80 may have an existing large-scale language model generate prompts to estimate and output character image information based on arbitrary text information input to the input unit 5, and then input the prompts generated by the existing large-scale language model to the image generation learning model (first learning model).

[0033] Figure 2 shows an example of estimating character image information. As shown in Figure 2, for example, the character image estimation unit 80 takes arbitrary text information, such as "The one who destroys all, born from the twilight," as input and estimates and outputs image information for a character (monster) named "Dark Lord Belzeforn." As the learning model for image generation (first learning model), an existing learning model for image generation (for example, Adobe's Firefly®) can be used.

[0034] Furthermore, if additional information indicating the worldview of the game card is input from the input unit 3 of the user device 2, the character image estimation unit 80 uses a learning model for image generation (first learning model) to estimate and output image information of the character to be generated for the game card, taking the arbitrary text information and additional information input to the input unit 5 as input. Specifically, the character image estimation unit 80 generates a prompt to estimate and output character image information based on the arbitrary text information and additional information input to the input unit 5, and inputs the generated prompt to the learning model for image generation (first learning model). As a result, the character image estimation unit 80 obtains character image information (character image information estimated from the arbitrary text information and additional information) as output from the learning model for image generation (first learning model).

[0035] The character details estimation unit 81 uses a second learning model different from the first learning model to take the character image information output from the character image estimation unit 80 as input and has the function of estimating and outputting the detailed information of that character. Specifically, the character details estimation unit 81 generates a prompt to estimate and output the detailed information of the character based on the character image information output from the character image estimation unit 80, and inputs the generated prompt to the second learning model. As a result, the character details estimation unit 81 obtains the detailed information of the character (detailed information of the character estimated from the character image information) as output from the second learning model.

[0036] Figure 3 shows an example of estimating detailed character information. As shown in Figure 3, for example, the character detailed information estimation unit 81 takes image information of a character (monster) called "Dark Lord Belzeforn" as input and estimates and outputs various detailed information about this character (monster) (e.g., monster name, hit points, physical attributes, physical characteristics, attack techniques, attack power, attack technique attributes, attack technique characteristics, etc.). As a second learning model, an existing large-scale language model (e.g., GPT) can be used.

[0037] The game card information generation unit 82 has the function of generating and outputting game card information for a character by combining the character image information output from the character image estimation unit 80 and the character details information output from the character details information estimation unit 81. Figure 4 shows an example of the generated game card information. For example, by combining the character image information shown in Figure 2 and the character details information shown in Figure 3, game card information as shown in Figure 4 is generated.

[0038] As described above, the generated game card information is transmitted from the output unit 6 to the user device 2. When game card information is generated, the text information input into the input unit 5 (pre-input text information) and the game card information generated by the game card information generation unit 82 based on that text information (pre-generated game card information) are stored in the memory unit 7 in correspondence with each other. For example, the text information "The one who destroys all born from the twilight" and the game card information for "Dark Lord Belzeforn" shown in Figure 5 are stored in the memory unit 7 in correspondence with each other.

[0039] The text matching determination unit 83 has the function of determining whether new text information entered into the input unit 5 matches previously entered text information stored in the storage unit 7. For example, if the text information "The one who is born from the twilight and destroys all" (previously entered text information) is newly entered from the input unit 5, the text matching determination unit 83 will determine that it matches the previously entered text information stored in the storage unit 7.

[0040] Then, if the text matching determination unit 83 determines that the new text information entered into the input unit 5 matches the previously entered text information stored in the storage unit 7, the game card information generation unit 82 outputs the generated game card information stored in the storage unit 7 in association with the newly entered text information. For example, if the text information "The one who destroys all born from the twilight" (previously entered text information) is newly entered from the input unit 5, the game card information generation unit 82 outputs the game card information (generated game card information) for "Dark Lord Belzeforn" shown in Figure 5. In this case, the character image estimation unit 80 does not perform the process of estimating and outputting character image information from arbitrary text information (previously entered text information), and therefore, the character details estimation unit 81 does not perform the process of estimating and outputting character details from character image information.

[0041] The operation of the game card generation system 1, configured as described above, will be explained with reference to the sequence diagram in Figure 5.

[0042] As shown in Figure 5, when generating a game card using the game card generation system 1 of this embodiment, first, arbitrary text information related to the game card is input into the user device 2 (S1). Additionally, additional information indicating the worldview of the game card to be generated is input into the user device 2 (S2). The arbitrary text information and additional information input into the user device 2 are transmitted to the game card generation system 1 (S3).

[0043] In the game card generation system 1, a determination is made as to whether the arbitrary text information transmitted from the user device 2 matches the input text information stored in the storage unit 7 (S4).

[0044] If the arbitrary text information transmitted from the user device 2 does not match the input text information stored in the memory unit 7, the image generation learning model (first learning model) is used to estimate the character's image information from the arbitrary text information and additional information (S5). Next, the second learning model is used to estimate the character's detailed information from the character's image information (S6). Then, the character's game card information is generated and output by combining the character's image information and the character's detailed information (S7). The generated game card information is stored in the memory unit 7 in association with the input arbitrary text information.

[0045] On the other hand, if any text information transmitted from the user device 2 matches the input text information stored in the storage unit 7, the game card information (generated game card information) stored in the storage unit 7 in association with that text information is output (S7).

[0046] The game card information generated from the arbitrary text information and additional information is transmitted from the game card generation system 1 to the user device 2 (S8) and displayed on the display unit 4 of the user device 2 (S9).

[0047] According to the game card generation system 1 of this embodiment, when arbitrary text information relating to a game card is input, a first learning model (a learning model for image generation) is used to generate character image information from the text information, and a second learning model (a learning model different from the first learning model) is used to generate detailed character information from the character image information. Then, the character image information generated by the first learning model and the detailed character information generated by the second learning model are combined to generate the game card information for that character. By generating the character's game card information (character image information and character detailed information) using two learning models (the first learning model and the second learning model) in this way, it becomes possible to improve the consistency between the character's image information and the character's detailed information compared to simply using a single learning model.

[0048] Furthermore, in this embodiment, when previously entered text information is entered again, the game card information is not regenerated based on the re-entered text information; instead, the previously generated game card information generated when that text information was previously entered is output. Depending on the learning model, if game card information is regenerated based on the re-entered text information, the previously generated game card information generated when that text information was previously entered may not necessarily be generated, which could lead to a decrease in reproducibility. A decrease in reproducibility could lead to a decrease in user satisfaction. According to the present invention, when previously entered text information is entered again, the previously generated game card information generated when that text information was previously entered is output, thus maintaining reproducibility.

[0049] Furthermore, in this embodiment, by inputting additional information that indicates the worldview of the game card in addition to the text information related to the game card, it is possible to generate character image information that reflects the worldview. According to the present invention, detailed information of the character is generated based on the character image information, and as a result, it is possible to generate detailed information of the character that reflects the worldview. In this way, it is possible to generate game card information of a character that reflects the worldview (character image information and character detailed information).

[0050] (modified version) Figure 5 is a block diagram showing a modified configuration of the game card generation system 1. In the modified configuration of Figure 5, the control unit 8 of the game card generation system 1 is equipped with an additional learning unit 84. Unless otherwise specified, the configuration and operation of the modified configuration of the game card generation system 1 are the same as those of the embodiment described above.

[0051] The additional learning unit 84 has the function of further training the first learning model by fine-tuning it using image information that represents the worldview of the game card. The training data used for additional learning (fine-tuning) can be stored in the memory unit 7 of the game card generation system 1. Known methods can be used for additional learning (fine-tuning).

[0052] In this case, the character image estimation unit 80 uses the newly trained first learning model to estimate and output image information of the character to be generated on the game card, taking the arbitrary text information input to the input unit 5 as input. Specifically, the character image estimation unit 80 generates a prompt to estimate and output character image information based on the arbitrary text information input to the input unit 5, and inputs the generated prompt to the newly trained image generation learning model (the newly trained first learning model). As a result, the character image estimation unit 80 obtains character image information (character image information estimated from the arbitrary text information) as output from the newly trained image generation learning model (the newly trained first learning model).

[0053] A modified version of the game card generation system 1 configured as described above will be explained with reference to the sequence diagram in Figure 7.

[0054] In this case, first, the game card generation system 1 performs additional learning (fine tuning) using image information that represents the worldview of the game card (S10). Then, when arbitrary text information related to the game card is input from the user device 2 (S11), the arbitrary text information and additional information input from the user device 2 are sent to the game card generation system 1 (S12). In this case, no additional information indicating the worldview of the game card to be generated is input from the user device 2.

[0055] In the game card generation system 1, a determination is made as to whether the arbitrary text information transmitted from the user device 2 matches the input text information stored in the storage unit 7 (S4).

[0056] If the arbitrary text information transmitted from the user device 2 does not match the input text information stored in the memory unit 7, the character image information is estimated from the arbitrary text information using the newly trained image generation model (first learning model) (S5). Next, the character's detailed information is estimated from the character's image information using the second learning model (S6). Then, the character's game card information is generated and output by combining the character's image information and the character's detailed information (S7). The generated game card information is stored in the memory unit 7 in association with the input arbitrary text information.

[0057] On the other hand, if any text information transmitted from the user device 2 matches the input text information stored in the storage unit 7, the game card information (generated game card information) stored in the storage unit 7 in association with that text information is output (S7).

[0058] The game card information generated from arbitrary text information is transmitted from the game card generation system 1 to the user device 2 (S8) and displayed on the display unit 4 of the user device 2 (S9).

[0059] Even with such a modified version of the game card generation system 1, the same effects and advantages as those of the embodiment described above can be achieved.

[0060] According to the modified game card generation system 1, by using image information that embodies the worldview of the game card and further training the first learning model through fine-tuning, it is possible to generate character image information that reflects the worldview. According to the present invention, detailed information of the character is generated based on the character image information, and as a result, it is possible to generate detailed information of the character that reflects the worldview. In this way, it is possible to generate game card information of a character that reflects the worldview (character image information and character detailed information).

[0061] Although embodiments of the present invention have been described above by example, the scope of the present invention is not limited to these, and modifications and alterations can be made within the scope described in the claims depending on the purpose. [Industrial applicability]

[0062] As described above, the game card generation system according to the present invention has the effect of improving the consistency between the character's image information and the character's detailed information when generating character game card information (character image information and character detailed information) from text information related to game cards, and is useful for generating competitive game cards and the like. [Explanation of Symbols]

[0063] 1. Game Card Generation System 2. User devices 3. Input section 4 Display section 5 Input section 6 Output section 7 Memory section 8 Control Unit 80 Character Image Estimation Section 81 Character Details Estimation Section 82 Game Card Information Generation Unit 83 Text Matching Unit 84 Additional Learning Section

Claims

1. An input section where arbitrary text information about the game card is entered, A character image estimation unit that uses a first learning model for image generation to estimate and output image information of the game card character to be generated, taking the arbitrary text information input to the input unit as input, A character detailed information estimation unit, using a second learning model different from the first learning model, takes the image information of the character output from the character image estimation unit as input, estimates and outputs detailed information of the character, A game card information generation unit combines the image information of the character output from the character image estimation unit and the detailed information of the character output from the character detailed information estimation unit to generate and output game card information of the character. A game card generation system equipped with the following features.

2. A storage unit that stores, in association with the input text information entered into the input unit and the generated game card information generated by the game card information generation unit based on the said text information, A text matching determination unit determines whether new text information entered into the input unit matches previously entered text information stored in the storage unit, Equipped with, The game card generation system according to claim 1, wherein the game card information generation unit outputs generated game card information stored in the storage unit in association with the input text information when the text matching determination unit determines that the new text information input to the input unit matches the input text information stored in the storage unit.

3. In addition to the arbitrary text information, the input unit receives additional information that indicates the worldview of the game card to be generated. The game card generation system according to claim 1, wherein the character image estimation unit uses the first learning model to estimate and output image information of the character of the game card to be generated, taking the arbitrary text information and the additional information input to the input unit as input.

4. The system includes an additional learning unit that further trains the first learning model by fine-tuning it using image information that represents the worldview of the game cards to be generated. The game card generation system according to claim 1, wherein the character image estimation unit uses the newly trained first learning model to estimate and output image information of the character of the game card to be generated, taking the arbitrary text information input to the input unit as input.

5. A method performed in a game card generation system, The aforementioned method, An input step in which arbitrary text information about the game card is entered, A character image output step that uses a first learning model for image generation to estimate and output image information of the game card character to be generated, taking the arbitrary text information input in the input step as input, A character detailed information output step that uses a second learning model different from the first learning model to estimate and output detailed information of the character, taking the image information of the character output in the character image output step as input, A game card information generation step that combines the image information of the character output in the character image output step and the detailed information of the character output in the character detailed information output step to generate and output game card information of the character, Methods that include...

6. A program executed by a game card generation system, The program is transmitted to the computer of the game card generation system. An input process where arbitrary text information about the game card is entered, A character image output process that uses a first learning model for image generation to estimate and output image information of the game card character to be generated, taking the arbitrary text information input in the input process as input, A character detailed information output process that uses a second learning model different from the first learning model to estimate and output detailed information of the character, taking the image information of the character output in the character image output process as input, A game card information generation process that combines the image information of the character output in the character image output process and the detailed information of the character output in the character detailed information output process to generate and output game card information for the character, A program that executes something.

Images