# OBS Recording Transcriber Dependencies # Core dependencies streamlit>=1.26.0 moviepy>=1.0.3 openai-whisper>=20231117 requests>=2.28.0 humanize>=4.6.0 # PyTorch ecosystem - DO NOT include here for Docker builds # These are installed separately with CUDA support in Dockerfile.gpu # For local installs: pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 # torchaudio >= 2.1.0 is REQUIRED for diarization to work properly # Transformers ecosystem transformers>=4.35.0 tokenizers>=0.14.0 # ML dependencies - use flexible versions for compatibility numpy>=1.24.0 scipy>=1.10.0 scikit-learn>=1.3.0 # Audio processing and ML models # speechbrain 1.0+ required for pyannote compatibility speechbrain>=1.0.0 pyannote.audio>=3.1.1 pytorch-lightning>=2.0.0 # Other dependencies iso639>=0.1.4 protobuf>=3.20.0,<5.0.0 matplotlib>=3.5.0 soundfile>=0.10.3 ffmpeg-python>=0.2.0 # Optional: Ollama Python client (uncomment to install) # ollama # Installation notes: # 1. For Windows users, you may need to install PyTorch separately: # pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 # # 2. For tokenizers issues, try installing Visual C++ Build Tools: # https://visualstudio.microsoft.com/visual-cpp-build-tools/ # # 3. For pyannote.audio, you'll need a HuggingFace token with access to: # https://huggingface.co/pyannote/speaker-diarization-3.0 # # 4. FFmpeg is required for audio processing: # Windows: https://www.gyan.dev/ffmpeg/builds/ # Mac: brew install ffmpeg # Linux: apt-get install ffmpeg