Text2Audio / requirements.txt
IFMedTechdemo's picture
split requirements for each model
c50bac9 verified
raw
history blame
1.42 kB
# Core dependencies (pinned for stability; updated for Oct 2025)
torch==2.9.0
torchaudio==2.9.0
torchvision==0.20.0 # Matches torch 2.9.0
numpy==2.0.0 # Stable 2.x; compatible with all (librosa, silero-vad, torch)
librosa==0.11.0 # Supports NumPy 2.x
soundfile==0.13.1
scipy>=1.11.0
pandas>=2.0.0
requests>=2.28.0
tqdm>=4.65.0
# HF Ecosystem (pinned to avoid known 2025 conflicts)
transformers==4.56.1 # Stable; avoids 4.57+ bugs with diffusers
diffusers==0.34.0 # Pre-0.35 TypeError; compatible with torch 2.9 & transformers 4.56
accelerate>=0.33.0 # Latest stable; works with Spaces
fsspec==2024.9.0 # Fixes gcsfs/diffusers conflicts
safetensors>=0.4.0
datasets>=2.14.0 # Loose for flexibility
# NeuTTS-Air specific (loosened where possible)
neucodec>=0.0.3
phonemizer==3.3.0
vector-quantize-pytorch==1.17.8
torchao>=0.3.0 # Updated for torch 2.9
torchtune>=0.3.0 # Loose
resemble-perth==1.0.1
# Kokoro TTS
kokoro>=0.2
bitsandbytes>=0.43.0 # Compatible with torch 2.9
snac>=0.1.0
pyttsx3>=3.1
gitpython>=3.1.0
# Chatterbox Multilingual (merged; conflicts resolved)
resampy==0.4.3
s3tokenizer>=0.1.0
omegaconf==2.3.0
silero-vad==6.0.0 # Latest; NumPy 2.x compatible via torch
conformer==0.3.2
# Optionals (uncomment for languages)
# spacy_pkuseg>=0.0.29 # Chinese
# pykakasi>=2.2.0 # Japanese
# russian-text-stresser @ git+https://github.com/Vuizur/add-stress-to-epub
# dicta-onnx>=0.1.0 # Hebrew