models/macros.py

# Text
NUM_TEXT_TOKENS = 2048

# Audio
NUM_AUDIO_TOKENS = 1024  # EnCodec RVQ bins
NUM_MEL_BINS = 100  # BigVGAN bigvgan_24khz_100band


# Speaker
NUM_SPEAKER_CLASSES = 4096
SPEAKER_EMBEDDING_DIM = 64