torch>=1.9.0
torchaudio>=0.9.0
huggingface_hub>=0.8.0
numpy>=1.19.0
transformers>=4.30.0
safetensors>=0.3.0
# phonemizer>=3.3.0   # only for phoneme_set="ph66" (also needs the espeak-ng system package)
librosa>=0.8.0
soundfile>=0.12.1
# torchcodec>=0.9.1   # only for audio formats soundfile cannot read (also needs FFmpeg)
matplotlib>=3.10.0