Method for synthesizing figure of virtual object, electronic device, and storage medium

1. A method for synthesizing a figure of a virtual object, comprising:

obtaining a figure image of the virtual object, and original face images corresponding to a speech segment, in which the figure image comprises a face of the virtual object;

extracting a first face key point of the face of the virtual object, and a second face key point of each of the original face images;

processing the first face key point based on a preset algorithm, to generate a position and posture information of a first three-dimensional 3D face corresponding to the face of the virtual object;

processing each second face key point based on a preset algorithm, to generate vertex information of a second 3D face corresponding to each original face image;

generating a target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face; and

synthesizing a speaking figure segment of the virtual object, corresponding to the speech segment, based on the figure image of the virtual object and each target face image.

2. The method of claim 1 , wherein, processing the first face key point based on the preset algorithm, to generate the position and the posture information of the first 3D face corresponding to the face of the virtual object, comprises:

performing coordinate transformation processing on the first face key point based on a preset transformation parameter, to generate a first 3D face mesh model; and

extracting the position and the posture information of the first 3D face corresponding to the face of the virtual object from the first 3D face mesh model, in which, the posture information comprises a pitch angle, a yaw angle, and a roll angle.

3. The method of claim 1 , wherein, processing each second face key point based on the preset algorithm, to generate the vertex information of the second 3D face corresponding to each original face image, comprises:

performing coordinate transformation processing on each second face key point based on a preset transformation parameter, to generate a second 3D face mesh model corresponding to each original face image; and

extracting the vertex information of the second 3D face from each second 3D face mesh model.

4. The method of claim 1 , wherein, generating the target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face, comprises:

performing projection transformation on the position and the posture information of the first 3D face and the vertex information of each second 3D face respectively based on a preset transformation algorithm, to generate a first target face image corresponding to each original face image.

5. The method of claim 4 , after the first target face image corresponding to each original face image is generated, further comprising:

processing each second face key point based on a preset algorithm, to generate a position and posture information of the second 3D face corresponding to each original face image;

obtaining texture information of a position corresponding to the position and the posture information of the second 3D face from each original face image based on a preset projection transformation algorithm; and

performing texture rendering on each first target face image based on the texture information of the corresponding position, to generate a second target face image corresponding to each original face image.

6. The method of claim 1 , wherein, synthesizing the speaking figure segment of the virtual object, corresponding to the speech segment, based on the figure image of the virtual object and each target face image, comprises:

generating a lower-half face mask corresponding to each target face image;

performing fusion processing on the figure image of the virtual object and each target face image based on a preset first fusion algorithm according to the lower-half face mask, to generate first fusion images; and

synthesizing the speaking figure segment of the virtual object corresponding to the speech segment based on the first fusion images.

7. The method of claim 6 , wherein, generating the lower-half face mask corresponding to each target face image, comprises:

performing edge detection on each target face image to obtain face edge points;

connecting the face edge points to form a closed region and filling a fixed value with the closed region to generate a full face mask; and

intercepting the lower-half face mask from the full face mask by taking a central key point of a nose bridge as reference.

8. The method of claim 6 , wherein, after the first fusion image corresponding to each original face image is generated, further comprises:

generating a mouth region mask corresponding to each target face image;

performing fusion processing on the figure image of the virtual object and each first fusion image based on a preset second fusion algorithm according to the mouth region mask, to generate second fusion images; and

synthesizing the speaking figure segment of the virtual object corresponding to the speech segment based on the second fusion images.

9. The method of claim 8 , wherein, generating the mouth region mask corresponding to each target face image, comprises:

extracting key points of an outer edge of a mouth of each target face image;

connecting the key points of the outer edge of the mouth to form a closed region and filling a fixed value with the closed region to generate an initial mask; and

performing Gaussian filtering on the initial mask to obtain the mouth region mask corresponding to each target face image.

10. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory is configured to store instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is caused to perform:

obtaining a figure image of a virtual object, and original face images corresponding to a speech segment, in which the figure image comprises a face of the virtual object;

extracting a first face key point of the face of the virtual object, and a second face key point of each of the original face images;

processing the first face key point based on a preset algorithm, to generate a position and posture information of a first three-dimensional 3D face corresponding to the face of the virtual object;

processing each second face key point based on a preset algorithm, to generate vertex information of a second 3D face corresponding to each original face image;

generating a target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face; and

synthesizing a speaking figure segment of the virtual object, corresponding to the speech segment, based on the figure image of the virtual object and each target face image.

11. The device of claim 10 , wherein, processing the first face key point based on the preset algorithm, to generate the position and the posture information of the first 3D face corresponding to the face of the virtual object, comprises:

performing coordinate transformation processing on the first face key point based on a preset transformation parameter, to generate a first 3D face mesh model; and

extracting the position and the posture information of the first 3D face corresponding to the face of the virtual object from the first 3D face mesh model, in which, the posture information comprises a pitch angle, a yaw angle, and a roll angle.

12. The device of claim 10 , wherein, processing each second face key point based on the preset algorithm, to generate the vertex information of the second 3D face corresponding to each original face image, comprises:

performing coordinate transformation processing on each second face key point based on a preset transformation parameter, to generate a second 3D face mesh model corresponding to each original face image; and

extracting the vertex information of the second 3D face from each second 3D face mesh model.

13. The device of claim 10 , wherein, generating the target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face, comprises:

performing projection transformation on the position and the posture information of the first 3D face and the vertex information of each second 3D face respectively based on a preset transformation algorithm, to generate a first target face image corresponding to each original face image.

14. The device of claim 13 , wherein when the instructions are executed by the at least one processor, the at least one processor is caused to perform:

after the first target face image corresponding to each original face image is generated, processing each second face key point based on a preset algorithm, to generate a position and posture information of the second 3D face corresponding to each original face image;

obtaining texture information of a position corresponding to the position and the posture information of the second 3D face from each original face image based on a preset projection transformation algorithm; and

performing texture rendering on each first target face image based on the texture information of the corresponding position, to generate a second target face image corresponding to each original face image.

15. The device of claim 10 , wherein, synthesizing the speaking figure segment of the virtual object, corresponding to the speech segment, based on the figure image of the virtual object and each target face image, comprises:

generating a lower-half face mask corresponding to each target face image;

performing fusion processing on the figure image of the virtual object and each target face image based on a preset first fusion algorithm according to the lower-half face mask, to generate first fusion images; and

synthesizing the speaking figure segment of the virtual object corresponding to the speech segment based on the first fusion images.

16. The device of claim 15 , wherein, generating the lower-half face mask corresponding to each target face image, comprises:

performing edge detection on each target face image to obtain face edge points;

connecting the face edge points to form a closed region and filling a fixed value with the closed region to generate a full face mask; and

intercepting the lower-half face mask from the full face mask by taking a central key point of a nose bridge as reference.

17. The device of claim 15 , wherein when the instructions are executed by the at least one processor, the at least one processor is caused to perform:

after the first fusion image corresponding to each original face image is generated, generating a mouth region mask corresponding to each target face image;

performing fusion processing on the figure image of the virtual object and each first fusion image based on a preset second fusion algorithm according to the mouth region mask, to generate second fusion images; and

synthesizing the speaking figure segment of the virtual object corresponding to the speech segment based on the second fusion images.

18. The device of claim 17 , wherein, generating the mouth region mask corresponding to each target face image, comprises:

extracting key points of an outer edge of a mouth of each target face image;

connecting the key points of the outer edge of the mouth to form a closed region and filling a fixed value with the closed region to generate an initial mask; and

performing Gaussian filtering on the initial mask to obtain the mouth region mask corresponding to each target face image.

19. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are configured to enable a computer to perform a method for synthesizing a figure of a virtual object, the method comprising:

obtaining a figure image of the virtual object, and original face images corresponding to a speech segment, in which the figure image comprises a face of the virtual object;

extracting a first face key point of the face of the virtual object, and a second face key point of each of the original face images;

processing the first face key point based on a preset algorithm, to generate a position and posture information of a first three-dimensional 3D face corresponding to the face of the virtual object;

processing each second face key point based on a preset algorithm, to generate vertex information of a second 3D face corresponding to each original face image;

generating a target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face; and

synthesizing a speaking figure segment of the virtual object, corresponding to the speech segment, based on the figure image of the virtual object and each target face image.

20. The non-transitory computer-readable storage medium of claim 19 , wherein, generating the target face image corresponding to each original face image based on the position and the posture information of the first 3D face and the vertex information of each second 3D face, comprises:

performing projection transformation on the position and the posture information of the first 3D face and the vertex information of each second 3D face respectively based on a preset transformation algorithm, to generate a first target face image corresponding to each original face image.