Nodes/ComfyUI-FLOAT_Optimized/FLOAT Apply Audio Projection (VA)
ComfyUI Node

FLOAT Apply Audio Projection (VA)

Applies the loaded audio projection layer to the features extracted from the Wav2Vec model. This final step projects the high-dimensional audio features down to the motion latent space, producing the final audio conditioning tensor (wa_latent).

By set-soft·Created about a year ago·Updated 12 months ago· 33
FLOAT Apply Audio Projection (VA)
  • wav2vec_features
  • projection_layer
  • wa_latent
CategoryFLOAT/Very Advanced

Inputs (2)

NameTypeDefaultDescription
wav2vec_featuresTORCH_TENSORThe batch of interpolated feature tensors output by the Wav2Vec feature extraction node.
projection_layerAUDIO_PROJECTION_LAYERThe loaded audio projection layer module.

Outputs (1)

NameTypeDescription
wa_latentTORCH_TENSOR