requirements.txt

# requirements.txt ── deps for Hugging Face Space
streamlit>=1.33         # front-end UI
transformers>=4.40      # Hugging Face pipelines
torch>=2.2              # core tensor library (CPU build auto-selected by HF)
soundfile               # write WAV audio
numpy                   # array ops for audio concat
pillow                  # image handling (PIL)
sentencepiece           # tokenizer back-end for some models
accelerate              # speeds up HF inference, optional but lightweight