Generating facial position data based on audio data
A computer-implemented method for generating a machine-learned model to generate facial position data based on audio data comprising training a conditional variational autoencoder having an encoder and decoder. The training comprises receiving a set of training data items, each training data item comprising a facial position descriptor and an audio descriptor; processing one or more of the training data items using the encoder to obtain distribution parameters; sampling a latent vector from a latent space distribution based on the distribution parameters; processing the latent vector and the audio descriptor using the decoder to obtain a facial position output; calculating a loss value based at least in part on a comparison of the facial position output and the facial position descriptor of at least one of the one or more training data items; and updating parameters of the conditional variational autoencoder based at least in part on the calculated loss value.
1. A computer-implemented method for generating facial animation data based on audio data comprising:
receiving an audio descriptor representative of audio data associated with a time step;
receiving a latent vector; and
processing the latent vector and the audio descriptor using a generative neural network model to output facial animation data, the facial animation data comprising either rig parameters or one or more values used to generate the rig parameters, wherein the rig parameters are configured to position a skeleton of a three-dimensional facial model in accordance with the latent vector and the audio descriptor.
2. The method of claim 1 , wherein the one or more values comprises a plurality of vertex positions as vertex coordinates or as displacements of the vertices from their positions in a template facial mesh, and wherein the one or more values are transformed to generate said rig parameters.
3. The method of claim 2 , wherein the one or more values are transformed to generate said rig parameters using a linear or non-linear transformation.
4. The method of claim 1 , comprising generating the latent vector based on one or more latent vectors associated with one or more time steps prior to the time step associated with the audio data.
5. The method of claim 1 , wherein the audio descriptor comprises a spectral representation of the audio data associated with the time step.
6. The method of claim 5 , wherein the spectral representation of the audio data comprises a plurality of mel-frequency cepstral coefficients.
7. The method of claim 1 , wherein the audio descriptor is an output of an encoder of a vector quantized variational autoencoder for the audio data associated with the time step.
8. The method of claim 1 , further comprising:
receiving one or more facial expression parameters; and
generating the latent vector based on the one or more facial expression parameters.
9. A system comprising:
at least one processor; and
a non-transitory computer-readable medium including executable instructions that when executed by the at least one processor cause the at least one processor to perform at least the following operations:
receive an audio descriptor representative of audio data associated with a time step;
receive a latent vector; and
process the latent vector and the audio descriptor using a generative neural network model to output facial animation data, the facial animation data comprising either rig parameters or one or more values used to generate the rig parameters, wherein the rig parameters are configured to position a skeleton of a three-dimensional facial model in accordance with the latent vector and the audio descriptor.
10. The system of claim 9 , wherein the one or more values comprises a plurality of vertex positions as vertex coordinates or as displacements of the vertices from their positions in a template facial mesh, and wherein the one or more values are transformed to generate said rig parameters.
11. The system of claim 10 , wherein the one or more values are transformed to generate said rig parameters using a linear or non-linear transformation.
12. The system of claim 9 , comprising generating the latent vector based on one or more latent vectors associated with one or more time steps prior to the time step associated with the audio data.
13. The system of claim 9 , wherein the audio descriptor comprises a spectral representation of the audio data associated with the time step.
14. A non-transitory computer-readable medium including executable instructions that when executed by one or more processors cause the one or more processors to perform at least the following operations:
receive an audio descriptor representative of audio data associated with a time step;
receive a latent vector; and
process the latent vector and the audio descriptor using a generative neural network model to output facial animation data, the facial animation data comprising either rig parameters or one or more values used to generate the rig parameters, wherein the rig parameters are configured to position a skeleton of a three-dimensional facial model in accordance with the latent vector and the audio descriptor.
15. The computer-readable medium of claim 14 , wherein the one or more values comprises a plurality of vertex positions as vertex coordinates or as displacements of the vertices from their positions in a template facial mesh, and wherein the one or more values are transformed to generate said rig parameters.
16. The computer-readable medium of claim 15 , wherein the one or more values are transformed to generate said rig parameters using a linear or non-linear transformation.
17. The computer-readable medium of claim 14 , comprising generating the latent vector based on one or more latent vectors associated with one or more time steps prior to the time step associated with the audio data.