A social media content generation system and method for a virtual idol

Abstract

The application discloses a kind of virtual idol's social media graphic content generation system and method, the system includes: information acquisition module, feature determination module, picture generation module, script generation module, result output module and database;Wherein, information acquisition module is used to obtain the virtual idol and target graphic theme information selected by user;Feature determination module is used to determine virtual idol feature, picture feature and script feature;Picture generation module is used to generate target picture after fusing according to the picture feature matching corresponding virtual idol's picture;Script generation module is used to determine script type, generate target script;Result output module is used to output target picture and target script;Database is used to store picture material and table data.The application can automatically generate the picture and script of virtual idol social media post by inputting text description, so as to improve the virtual idol social media graphic content creation efficiency.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and information processing technology, specifically to a system and method for generating social media graphic content for virtual idols. Background Technology

[0002] Virtual idols are virtual characters created artificially through computer graphics and other means. Early virtual idols were concentrated in the realm of anime and manga, but in recent years, with the continuous maturation of computer technology, many more human-like and realistic virtual idols have emerged. Currently, numerous virtual idols with diverse appearances and rich personalities exist on domestic and international social media platforms. They closely resemble real people in appearance, share their daily lives and social interactions like real bloggers, and participate in various commercial activities and product promotions.

[0003] Currently, the images and text in virtual idol social media posts are all manually created by the teams behind them. This creative model suffers from limited resources, long lead times, high costs, and low efficiency. Especially in image creation, the images in virtual idol social media posts primarily use real-life photos of models, which are then stitched together with the virtual idol's head model using post-processing techniques to create the final image. This process is limited by the team's resources, such as real-life models and background scenes, and requires significant human and material resources. Furthermore, with the increasing number of virtual idols and the frequent updates to social media content, the problems of limited resources, long lead times, high costs, and low efficiency faced by manual creation will become more pronounced, making it difficult to maintain a high-frequency, continuous output of text and image content. Summary of the Invention

[0004] The present invention addresses the shortcomings of the existing technology by proposing a system and method for generating social media graphic content for virtual idols. This system aims to automatically generate images and text for relevant social posts by inputting text descriptions, thereby improving the efficiency of creating social media graphic content for virtual idols.

[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0006] The present invention provides a social media graphic content generation system for virtual idols, characterized by comprising: an information acquisition module, a feature determination module, an image generation module, a text generation module, a result output module, and a database; wherein, the database stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords;

[0007] The information acquisition module includes: a virtual idol information acquisition unit and a graphic and text theme information acquisition unit;

[0008] The feature determination module includes: a virtual idol feature determination unit, an image feature determination unit, and a text feature determination unit;

[0009] The image generation module includes: an image material matching unit, an image material determination unit, and an image material fusion unit;

[0010] The copy generation module includes: a copy type determination unit and a copy content generation unit;

[0011] The virtual idol information acquisition unit acquires the name of the virtual idol selected by the user from the list of virtual idols;

[0012] The image and text topic information acquisition unit acquires the target image and text topic information input by the user;

[0013] The virtual idol feature determination unit extracts the corresponding target virtual idol's ID, tag, and character setting information from the virtual idol list based on the selected virtual idol name;

[0014] The image feature determination unit determines the expression type, clothing type, action type, and background scene type of the target virtual idol based on the input target image and text theme information;

[0015] The text feature determination unit determines the theme and language requirements of the target text based on the input target text and image theme information;

[0016] The image material matching unit includes: a first matching subunit, a second matching subunit, a third matching subunit, and a fourth matching subunit;

[0017] The first matching subunit matches the character expression material library and image material library of the target virtual idol corresponding to the number from the database;

[0018] The second matching subunit matches a set of matching virtual idol expression images from the target virtual idol's character expression material library based on the target virtual idol's expression type, and forms a first pre-selection set;

[0019] The third matching subunit matches a set of images of the target virtual idol's clothing type from the target virtual idol's image material library, and then matches a set of images of the virtual idol that match the clothing type from the set of images of the target virtual idol's actions, thus forming a second pre-selection set.

[0020] The fourth matching subunit matches a set of background scene images that match the background scene type from the background scene material library, and forms a third pre-selection set;

[0021] The image material determination unit includes: a first determination subunit, a second determination subunit, and a third determination subunit;

[0022] The first determining subunit is used to obtain the target virtual idol emoticon image selected by the user from the first pre-selection set;

[0023] The second determining subunit is used to obtain the target virtual idol image image selected by the user from the second pre-selection set;

[0024] The third determining subunit is used to obtain the target background scene image selected by the user from the third preselection set;

[0025] The image material fusion unit includes: a first fusion subunit and a second fusion subunit;

[0026] The first fusion subunit uses the target virtual idol image as the base image and fuses the target virtual idol expression image with the facial image in the target virtual idol image to obtain the target virtual idol image;

[0027] The second fusion subunit uses the target background scene image as the base image, makes the background of the target virtual idol image transparent to obtain the virtual idol character, and then merges it with the target background scene image to obtain the target image, which is then displayed through the result output module.

[0028] The copy type determination unit calculates the similarity between the target copy and the subject words of each type of copy in the copy type and subject word list, based on the theme of the target copy, and then takes the copy type corresponding to the subject word with the highest similarity as the target copy type.

[0029] The text content generation unit generates target text based on the virtual idol's tags, character setting information, and language requirements of the target text, and displays it through the result output module.

[0030] The present invention provides a method for generating social media graphic content for virtual idols, characterized by the following steps:

[0031] Step 1: Build a database that stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords;

[0032] Step 2: Obtain the name of the virtual idol selected by the user from the list of virtual idols and the target image and text theme information entered by the user;

[0033] Step 3: Based on the selected virtual idol name, extract the corresponding target virtual idol's ID, tag, and character setting information from the virtual idol list;

[0034] Step 4: Determine the target virtual idol's expression type, clothing type, action type, and background scene type based on the input target image and text theme information;

[0035] Step 5: Determine the theme and language requirements of the target text based on the input target image and text theme information;

[0036] Step 6: Based on the target virtual idol number, match the character expression material library and image material library of the target virtual idol corresponding to the corresponding number from the database;

[0037] Step 7: Based on the target virtual idol's expression type, match the set of matching virtual idol expression images from the target virtual idol's character expression material library, and form the first pre-selection set;

[0038] Step 8: Based on the clothing type of the target virtual idol, match the matching clothing type image set from the target virtual idol's image material library, and then match the matching virtual idol image image set from the matching clothing type image set based on the target virtual idol's action type, thus forming a second pre-selection set;

[0039] Step 9: Based on the background scene type, match a set of background scene images that match the background scene material library and form a third pre-selection set;

[0040] Step 10: Obtain the emoticon images of the target virtual idol selected by the user from the first pre-selection set;

[0041] Step 11: Obtain the image of the target virtual idol selected by the user from the second pre-selection set;

[0042] Step 12: Obtain the target background scene image selected by the user from the third pre-selection set;

[0043] Step 13: Using the target virtual idol image as the base image, fuse the target virtual idol expression image with the facial image in the target virtual idol image image. This includes: aligning the facial proportions and angles of the target virtual idol expression image and the target virtual idol image image; then detecting the facial key points in the target virtual idol expression image and the target virtual idol image image respectively; and using the triangular fusion method to fuse the two sets of detected facial key points to obtain the fused facial image. Finally, blurring the edges of the fused facial image to obtain the target virtual idol image.

[0044] Step 14: Using the target background scene image as the base image, make the background of the target virtual idol image transparent to obtain the virtual idol character, and then merge it with the target background scene image. This includes: first, determining the position and size of the virtual idol character in the background scene, and processing the position of the virtual idol character in the background scene and the occlusion relationship between the objects at the corresponding positions, so that the virtual idol character appears in the background scene. Then, using image harmonization methods, the color and lighting of the virtual idol character and the background scene are adjusted, and corresponding shadows or reflections are added to the virtual idol character according to the lighting of the background scene, so as to obtain the target image.

[0045] Step 15: Based on the theme of the target copy, calculate its similarity with the theme words of each type of copy in the copy type and theme word list, and then take the copy type corresponding to the theme word with the highest similarity as the target copy type;

[0046] Step 16: Generate the target text based on the virtual idol's tags, character setting information, and the language requirements of the target text;

[0047] Step 17: Output the target image and target text to the user interface.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0049] 1. The virtual idol social media image and text content generation system and method provided by the present invention can effectively utilize the image data generated during the virtual idol modeling process. With the support of existing databases and related image fusion algorithms and text generation models, it realizes the automated generation of virtual idol social media posts, reduces the manual cost of image processing and text writing in the creation process of virtual idol social media posts, and improves creation efficiency.

[0050] 2. The feature determination module in the virtual idol social media graphic content generation system provided by this invention is used to analyze the target graphic description information input by the user, and extract effective keywords about the target virtual idol, target image and target text as the basis for subsequent generation, thereby ensuring the accuracy and reliability of the content generated by the system.

[0051] 3. The image generation module in the virtual idol social media graphic content generation system provided by the present invention is used to match images of expressions, actions, clothing and background scenes that conform to the image characteristics from an existing material library. At the same time, it obtains the target image selected by the user from the matching results and merges them. While ensuring that each feature can be accurately displayed, it also absorbs the user's judgment on the image elements and overall style, thus ensuring the authenticity and naturalness of the merged image.

[0052] 4. The text generation module in the virtual idol social media graphic content generation system provided by this invention is used to determine the text type based on the target text theme information, and then generate the target text using a text generator of the corresponding type based on the virtual idol tags, character settings and target text language requirements. It takes into account the differences in text structure in different application scenarios, and combines the character settings of the virtual idol to ensure the fluency of the text while strengthening the personal characteristics of the virtual idol. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0054] Figure 2 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0055] In this embodiment, as Figure 1 As shown, a social media graphic content generation system for virtual idols includes: an information acquisition module, a feature determination module, an image generation module, a text generation module, a result output module, and a database; wherein, the database stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords;

[0056] The virtual idol list includes the virtual idol's name, number, tags, and character setting information; each virtual idol in the list has a corresponding character expression material library and image material library. The character expression material library contains virtual idol expression images with corresponding expression tags, and the image material library contains virtual idol image images with corresponding clothing type tags and action tags; the background scene material library contains scene images with corresponding scene tags, including: indoor / outdoor, city, street, and building;

[0057] The formation of the text type and its keyword list includes: first, acquiring social media post texts and building a text database; then, labeling some post texts with text types and adding text type tags; building a training set for the text classification model and training the model; using the trained text classification model to determine the text type of the remaining posts and adding text type tags; and performing LDA topic extraction on all post texts of the same type to obtain the keywords for each type of text.

[0058] The information acquisition module includes: a virtual idol information acquisition unit and a graphic and text theme information acquisition unit;

[0059] The feature determination module includes: a virtual idol feature determination unit, an image feature determination unit, and a text feature determination unit;

[0060] The image generation module includes: an image material matching unit, an image material determination unit, and an image material fusion unit;

[0061] The copywriting generation module includes: a copywriting type determination unit and a copywriting content generation unit;

[0062] The virtual idol information acquisition unit acquires the name of the virtual idol selected by the user from the list of virtual idols, for example: the selected virtual idol is AYAYI;

[0063] The image and text theme information acquisition unit acquires the target image and text theme information input by the user. The image and text theme information may include event, location, time and related details. For example, it can generate an image and text post about sharing daily life, where a person is wearing casual clothes and reading a book in a bookstore, sharing the book "Nromal People". The theme of the book is a love story about two people from broken families who save each other.

[0064] The virtual idol feature determination unit extracts the corresponding target virtual idol's ID, tag, and character setting information from the virtual idol list based on the selected virtual idol name. In this embodiment, based on the name AYAYI, the corresponding information extracted from the virtual idol list is ID VI0032, tag is fashion model, and character setting is Tmall Super Brand Digital Manager.

[0065] The image feature determination unit determines the expression type, clothing type, action type, and background scene type of the target virtual idol based on the input target image and text theme information. In this embodiment, the possible analysis results are: expression type is natural, smiling, or thinking; clothing type is daily / casual; action type is reading; and background scene type is bookstore or bookshelf.

[0066] The text feature determination unit determines the theme and language requirements of the target text based on the input target image and text theme information. In this embodiment, the possible analysis results are: the theme is book sharing "Nromal People", a love story of two people from broken families saving each other, and the language requirement is literary or lyrical.

[0067] The image material matching unit includes: a first matching subunit, a second matching subunit, a third matching subunit, and a fourth matching subunit;

[0068] The first matching subunit matches the character expression material library and image material library of the target virtual idol from the database according to the target virtual idol's number. In this embodiment, the corresponding character expression material library VI0032_FACE and image material library VI0032_LOOK are matched according to the number VI0032.

[0069] The second matching subunit matches a set of virtual idol expression images that match the expression type of the target virtual idol from the character expression material library of the target virtual idol, and forms a first pre-selection set. In this embodiment, images with the expression tags of natural, smiling or thinking are matched from the VI0032_FACE library, and all matched images are included in the first pre-selection set.

[0070] The third matching subunit matches a set of images with matching clothing types from the target virtual idol's image material library based on the target virtual idol's clothing type. Then, based on the target virtual idol's action type, it matches a set of images of the virtual idol with matching clothing types from the set of matching images, and forms the second pre-selection set. For example, in the example, images with clothing tags of "daily" or "casual" and action type of "reading" are matched from the VI0032_LOOK library. All matched images are included in the second pre-selection set.

[0071] The fourth matching subunit matches a set of background scene images that match the background scene type from the background scene material library and forms the third pre-selection set. In this embodiment, images with the scene tag of indoor bookshelf or bookstore are matched from the background scene material library, and all matched images are put into the third pre-selection set.

[0072] The image material determination unit includes a first determination subunit, a second determination subunit, and a third determination subunit. Specifically, each pre-selection set includes multiple images that meet the characteristics. The tones and styles of the elements in different images are different. This step can obtain the images that the user selects from each pre-selection set that are relatively matched in style and tone, as the basis for generating the target image.

[0073] The first determining subunit is used to obtain the target virtual idol emoticon image selected by the user from the first pre-selection set;

[0074] The second determining subunit is used to obtain the target virtual idol image image selected by the user from the second pre-selection set;

[0075] The third determining subunit is used to obtain the target background scene image selected by the user from the third preselection set;

[0076] The image fusion unit includes: a first fusion subunit and a second fusion subunit;

[0077] The first fusion subunit uses the target virtual idol image as the base image and merges the target virtual idol expression image with the facial image in the target virtual idol image to obtain the target virtual idol image;

[0078] The second fusion subunit uses the target background scene image as the base image, makes the background in the target virtual idol image transparent to obtain the virtual idol character, and then merges it with the target background scene image to obtain the target image, which is then displayed through the result output module.

[0079] The copy type determination unit calculates the similarity between the target copy's theme and the theme words of each type of copy in the copy type and theme word list, and then selects the copy type corresponding to the theme word with the highest similarity as the target copy type.

[0080] The copywriting generation unit generates target copywriting based on the virtual idol's tags, character setting information, and the language requirements of the target copywriting, and displays the results through the output module.

[0081] In this embodiment, as Figure 2 As shown, a method for generating social media image and text content for virtual idols proceeds through the following steps:

[0082] Step 1: Build a database that stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords;

[0083] Step 2: Obtain the name of the virtual idol selected by the user from the list of virtual idols and the target image and text theme information entered by the user;

[0084] Step 3: Based on the selected virtual idol name, extract the corresponding target virtual idol's ID, tags, and character setting information from the virtual idol list;

[0085] Step 4: Determine the target virtual idol's expression type, clothing type, action type, and background scene type based on the input target image and text theme information;

[0086] Step 5: Determine the theme and language requirements of the target text based on the input target image and text theme information;

[0087] Step 6: Based on the target virtual idol's ID, match the corresponding virtual idol's character expression material library and image material library from the database;

[0088] Step 7: Based on the target virtual idol's expression type, match the set of virtual idol expression images that match the target virtual idol's character expression material library, and form the first pre-selection set;

[0089] Step 8: Based on the clothing type of the target virtual idol, match the matching clothing type image set from the target virtual idol's image material library, and then match the matching virtual idol image set from the matching clothing type image set based on the target virtual idol's action type, thus forming the second pre-selection set;

[0090] Step 9: Based on the background scene type, match the matching background scene image set from the background scene material library and form the third pre-selection set;

[0091] Step 10: Obtain the emoticon images of the target virtual idol selected by the user from the first pre-selection set;

[0092] Step 11: Obtain the image of the target virtual idol selected by the user from the second pre-selection set;

[0093] Step 12: Obtain the target background scene image selected by the user from the third pre-selection set;

[0094] Step 13: Using the target virtual idol image as the base image, fuse the target virtual idol expression image with the facial image in the target virtual idol image image. This includes: aligning the facial proportions and angles of the target virtual idol expression image and the target virtual idol image image; then detecting the facial key points in the target virtual idol expression image and the target virtual idol image image respectively; and using the triangular fusion method to fuse the two sets of detected facial key points to obtain the fused facial image. Finally, blurring the edges of the fused facial image to obtain the target virtual idol image.

[0095] Step 14: Using the target background scene image as the base image, make the background of the target virtual idol image transparent to obtain the virtual idol character. Then, merge it with the target background scene image. This includes: first, determining the position and size of the virtual idol character in the background scene, and processing the position of the virtual idol character in the background scene, as well as the occlusion relationship between the objects in the corresponding position, so that the virtual idol character appears in the background scene. Then, using image harmonization methods, adjust the color and lighting of the virtual idol character and the background scene, and add corresponding shadows or reflections to the virtual idol character according to the lighting of the background scene to obtain the target image.

[0096] Step 15: Based on the theme of the target copy, calculate its similarity to the theme words of each type of copy in the copy type and theme word list, and then take the copy type corresponding to the theme word with the highest similarity as the target copy type.

[0097] Step 16: Generate the target text based on the virtual idol's tags, character setting information, and the language requirements of the target text;

[0098] Step 17: Output the target image and target text to the user interface.

Claims

1. A social media graphic content generation system for virtual idols, characterized in that, include: The system includes an information acquisition module, a feature determination module, an image generation module, a text generation module, a result output module, and a database. The database stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords. The information acquisition module includes: a virtual idol information acquisition unit and a graphic and text theme information acquisition unit; The feature determination module includes: a virtual idol feature determination unit, an image feature determination unit, and a text feature determination unit; The image generation module includes: an image material matching unit, an image material determination unit, and an image material fusion unit; The copy generation module includes: a copy type determination unit and a copy content generation unit; The virtual idol information acquisition unit acquires the name of the virtual idol selected by the user from the list of virtual idols; The image and text topic information acquisition unit acquires the target image and text topic information input by the user; The virtual idol feature determination unit extracts the corresponding target virtual idol's ID, tag, and character setting information from the virtual idol list based on the selected virtual idol name; The image feature determination unit determines the expression type, clothing type, action type, and background scene type of the target virtual idol based on the input target image and text theme information; The text feature determination unit determines the theme and language requirements of the target text based on the input target text and image theme information; The image material matching unit includes: a first matching subunit, a second matching subunit, a third matching subunit, and a fourth matching subunit; The first matching subunit matches the character expression material library and image material library of the target virtual idol corresponding to the number from the database; The second matching subunit matches a set of matching virtual idol expression images from the target virtual idol's character expression material library based on the target virtual idol's expression type, and forms a first pre-selection set; The third matching subunit matches a set of images of the target virtual idol's clothing type from the target virtual idol's image material library, and then matches a set of images of the virtual idol that match the clothing type from the set of images of the target virtual idol's actions, thus forming a second pre-selection set. The fourth matching subunit matches a set of background scene images that match the background scene type from the background scene material library, and forms a third pre-selection set; The image material determination unit includes: a first determination subunit, a second determination subunit, and a third determination subunit; The first determining subunit is used to obtain the target virtual idol emoticon image selected by the user from the first pre-selection set; The second determining subunit is used to obtain the target virtual idol image image selected by the user from the second pre-selection set; The third determining subunit is used to obtain the target background scene image selected by the user from the third preselection set; The image material fusion unit includes: a first fusion subunit and a second fusion subunit; The first fusion subunit uses the target virtual idol image as the base image and fuses the target virtual idol expression image with the facial image in the target virtual idol image to obtain the target virtual idol image; The second fusion subunit uses the target background scene image as the base image, makes the background of the target virtual idol image transparent to obtain the virtual idol character, and then merges it with the target background scene image to obtain the target image, which is then displayed through the result output module. The copy type determination unit calculates the similarity between the target copy and the subject words of each type of copy in the copy type and subject word list, based on the theme of the target copy, and then takes the copy type corresponding to the subject word with the highest similarity as the target copy type. The text content generation unit generates target text based on the virtual idol's tags, character setting information, and language requirements of the target text, and displays it through the result output module.

2. A method for generating social media graphic content for virtual idols, characterized in that, The procedure is as follows: Step 1: Build a database that stores a list of virtual idols, a library of virtual idol character expression materials and image materials, a library of background scene materials, and a list of text types and their keywords; Step 2: Obtain the name of the virtual idol selected by the user from the list of virtual idols and the target image and text theme information entered by the user; Step 3: Based on the selected virtual idol name, extract the corresponding target virtual idol's ID, tag, and character setting information from the virtual idol list; Step 4: Determine the target virtual idol's expression type, clothing type, action type, and background scene type based on the input target image and text theme information; Step 5: Determine the theme and language requirements of the target text based on the input target image and text theme information; Step 6: Based on the target virtual idol number, match the character expression material library and image material library of the target virtual idol corresponding to the corresponding number from the database; Step 7: Based on the target virtual idol's expression type, match the set of matching virtual idol expression images from the target virtual idol's character expression material library, and form the first pre-selection set; Step 8: Based on the clothing type of the target virtual idol, match the matching clothing type image set from the target virtual idol's image material library, and then match the matching virtual idol image image set from the matching clothing type image set based on the target virtual idol's action type, thus forming a second pre-selection set; Step 9: Based on the background scene type, match a set of background scene images that match the background scene material library and form a third pre-selection set; Step 10: Obtain the emoticon images of the target virtual idol selected by the user from the first pre-selection set; Step 11: Obtain the image of the target virtual idol selected by the user from the second pre-selection set; Step 12: Obtain the target background scene image selected by the user from the third pre-selection set; Step 13: Using the target virtual idol image as the base image, merge the target virtual idol expression image with the facial image in the target virtual idol image. This includes: aligning the facial proportions and angles of the target virtual idol expression image and the target virtual idol image; then detecting the facial key points in the target virtual idol expression image and the target virtual idol image respectively; and using the triangular fusion method to merge the two sets of detected facial key points to obtain the merged facial image. Finally, blurring the edges of the merged facial image to obtain the target virtual idol image. Step 14: Using the target background scene image as the base image, make the background of the target virtual idol image transparent to obtain the virtual idol character, and then merge it with the target background scene image. This includes: first, determining the position and size of the virtual idol character in the background scene, and processing the position of the virtual idol character in the background scene and the occlusion relationship between the objects at the corresponding positions, so that the virtual idol character appears in the background scene. Then, using image harmonization methods, the color and lighting of the virtual idol character and the background scene are adjusted, and corresponding shadows or reflections are added to the virtual idol character according to the lighting of the background scene, so as to obtain the target image. Step 15: Based on the theme of the target copy, calculate its similarity with the theme words of each type of copy in the copy type and theme word list, and then take the copy type corresponding to the theme word with the highest similarity as the target copy type; Step 16: Generate the target text based on the virtual idol's tags, character setting information, and the language requirements of the target text; Step 17: Output the target image and target text to the user interface.

Images