Three-dimensional face animation from speech

1. A computer-implemented method, comprising:

identifying, from an audio capture of a subject, an audio-correlated facial feature;

generating a first mesh for a lower portion of a face of the subject, based on the audio-correlated facial feature;

identifying an expression-like facial feature of the subject;

generating a second mesh for an upper portion of the face of the subject based on the expression-like facial feature;

forming a synthesized mesh with the first mesh and the second mesh;

generating a three-dimensional model of the face of the subject with the synthesized mesh based on a ground truth image of the subject; and

providing the three-dimensional model of the face of the subject to a display in a client device running an immersive reality application that includes the subject.

2. The computer-implemented method of claim 1 , further comprising receiving the audio capture of the subject from a virtual reality headset.

3. The computer-implemented method of claim 1 , wherein identifying the audio-correlated facial feature comprises identifying an intensity and a frequency of the audio capture from the subject and correlating an amplitude and a frequency of an audio waveform with a geometry of the lower portion of the face of the subject.

4. The computer-implemented method of claim 1 , wherein generating the first mesh comprises including an eyeblink or an eyebrow motion of the subject.

5. The computer-implemented method of claim 1 , wherein identifying the expression-like facial feature of the subject comprises stochastically selecting the expression-like facial feature based on a prior sampling of multiple facial expressions of the subject.

6. The computer-implemented method of claim 1 , wherein identifying the expression-like facial feature of the subject comprises correlating an upper facial feature with a speech feature from the audio capture of the subject.

7. The computer-implemented method of claim 1 , wherein identifying the expression-like facial feature of the subject comprises using a stochastic sampling of multiple facial expressions collected during a training session of a second subject reading a text or in a conversation.

8. The computer-implemented method of claim 1 , wherein generating the second mesh comprises accessing a three-dimensional model of the face of the subject with a neutral expression.

9. The computer-implemented method of claim 1 , wherein forming the synthesized mesh comprises merging a lip shape in the first mesh into an eye closure in the second mesh continuously across the face of the subject.

10. The computer-implemented method of claim 1 , further comprising receiving the audio capture of the subject together with an image capture of the face of the subject, and generating the second mesh comprises using the image capture.

11. A system, comprising:

one or more processors; and

a memory storing instructions which, when executed by the one or more processors, cause the system to:

identify, from an audio capture of a subject, an audio-correlated facial feature;

generate a first mesh for a lower portion of a face of the subject, based on the audio-correlated facial feature;

identify an expression-like facial feature of the subject;

generate a second mesh for an upper portion of the face of the subject based on the expression-like facial feature;

form a synthesized mesh with the first mesh and the second mesh;

generate a three-dimensional model of the face of the subject with the synthesized mesh based on a ground truth image of the subject; and

provide the three-dimensional model of the face of the subject to a display in a client device running an immersive reality application that includes the subject.

12. The system of claim 11 , wherein the one or more processors further execute instructions to receive the audio capture of the subject from a virtual reality headset.

13. The system of claim 11 , wherein to identify the expression-like facial feature of the subject the one or more processors execute instructions to stochastically select the expression-like facial feature based on a prior sampling of multiple facial expressions of the subject.

14. The system of claim 11 , wherein to identify the expression-like facial feature of the subject the one or more processors execute instructions to correlate an upper facial feature with a speech feature from the audio capture of the subject.

15. The system of claim 11 , wherein to identify the expression-like facial feature of the subject the one or more processors execute instructions to use a stochastic sampling of multiple facial expressions collected during a training session of a second subject reading a text or in a conversation.

16. A computer-implemented method, comprising:

determining a first correlation value for a facial feature based on an audio waveform from a first subject;

generating a plurality of meshes configured for different portions of a human face, based on the facial feature and at least the first correlation value;

updating the first correlation value based on a difference between at least one of the plurality of meshes and a ground truth image of the first subject; and

providing a three-dimensional model of the human face animated by speech to an immersive reality application accessed by a client device based on the difference between the least one of the plurality of meshes first mesh and the ground truth image of the first subject.

17. The computer-implemented method of claim 16 , further comprising:

determining a second correlation value for an upper facial feature;

generating a mesh for an upper portion of the human face based on the upper facial feature and the second correlation value;

forming a synthesized mesh with the mesh; and

forming the three-dimensional model of the human face animated by speech with the synthesized mesh.

18. The computer-implemented method of claim 16 , wherein determining the first correlation value for the facial feature comprises identifying the facial feature based on an intensity and a frequency of the audio waveform.

19. The computer-implemented method of claim 16 , further comprising determining a loss value of the plurality of meshes based on the ground truth image of the first subject.

20. The computer-implemented method of claim 16 , further comprising updating the first correlation value for the facial feature based on a second audio waveform from a second subject.