mirror of
https://github.com/ThePlasmak/faster-whisper.git
synced 2026-09-19 07:38:23 +08:00
ThePlasmak
28b64fdcae
- setup.sh --check: replace full pyannote.audio import with importlib.util.find_spec + metadata version lookup; avoids 30-60s CUDA init, check now completes in ~12s - skill.json: move ffmpeg from requires.bins to optionalBins; ffmpeg not needed for basic transcription since PyAV handles decoding (only needed for --burn-in, --normalize, --denoise, --channel, --export-speakers) - CHANGELOG.md + SKILL.md: document both fixes
348 lines
12 KiB
Bash
Executable File
348 lines
12 KiB
Bash
Executable File
#!/usr/bin/env bash
|
||
# faster-whisper skill setup
|
||
# Creates venv and installs dependencies (with GPU support where available)
|
||
#
|
||
# Usage:
|
||
# ./setup.sh # Base install
|
||
# ./setup.sh --diarize # Base install + speaker diarization (pyannote.audio)
|
||
|
||
set -e
|
||
|
||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||
VENV_DIR="$SCRIPT_DIR/.venv"
|
||
INSTALL_DIARIZE=false
|
||
|
||
# Parse arguments
|
||
for arg in "$@"; do
|
||
case "$arg" in
|
||
--diarize) INSTALL_DIARIZE=true ;;
|
||
--update)
|
||
# Upgrade faster-whisper in existing venv without full reinstall
|
||
if [ ! -d "$VENV_DIR" ]; then
|
||
echo "❌ No venv found at $VENV_DIR — run ./setup.sh first"
|
||
exit 1
|
||
fi
|
||
if command -v uv &> /dev/null; then
|
||
uv pip install --python "$VENV_DIR/bin/python" --upgrade faster-whisper
|
||
else
|
||
"$VENV_DIR/bin/pip" install --upgrade faster-whisper
|
||
fi
|
||
echo "✅ faster-whisper updated"
|
||
"$VENV_DIR/bin/python" -c "import faster_whisper; print(f'Version: {faster_whisper.__version__}')"
|
||
exit 0
|
||
;;
|
||
--check)
|
||
# Quick system check: GPU, Python, ffmpeg, venv, faster-whisper, yt-dlp, pyannote
|
||
echo "🔍 faster-whisper skill system check"
|
||
echo ""
|
||
|
||
# Python
|
||
if command -v python3 &>/dev/null; then
|
||
PY_VER=$(python3 -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}")')
|
||
echo "✅ Python: $PY_VER"
|
||
else
|
||
echo "❌ Python: not found"
|
||
fi
|
||
|
||
# ffmpeg
|
||
if command -v ffmpeg &>/dev/null; then
|
||
FF_VER=$(ffmpeg -version 2>&1 | head -1 | awk '{print $3}')
|
||
echo "✅ ffmpeg: $FF_VER"
|
||
else
|
||
echo "⚠️ ffmpeg: not found (needed for --normalize, --denoise, --burn-in)"
|
||
fi
|
||
|
||
# GPU/CUDA
|
||
NVIDIA_SMI_CHECK=""
|
||
if command -v nvidia-smi &>/dev/null; then
|
||
NVIDIA_SMI_CHECK="nvidia-smi"
|
||
elif grep -qi microsoft /proc/version 2>/dev/null; then
|
||
for wsl_smi in /usr/lib/wsl/lib/nvidia-smi /usr/lib/wsl/drivers/*/nvidia-smi; do
|
||
[ -f "$wsl_smi" ] && NVIDIA_SMI_CHECK="$wsl_smi" && break
|
||
done
|
||
fi
|
||
if [ -n "$NVIDIA_SMI_CHECK" ]; then
|
||
GPU_CHECK=$("$NVIDIA_SMI_CHECK" --query-gpu=name --format=csv,noheader 2>/dev/null | head -1)
|
||
DRV_CHECK=$("$NVIDIA_SMI_CHECK" --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1)
|
||
if [ -n "$GPU_CHECK" ]; then
|
||
echo "✅ GPU: $GPU_CHECK (driver $DRV_CHECK)"
|
||
else
|
||
echo "⚠️ GPU: nvidia-smi found but no GPU reported"
|
||
fi
|
||
else
|
||
echo "⚠️ GPU: no NVIDIA GPU detected (CPU mode)"
|
||
fi
|
||
|
||
# venv
|
||
if [ -d "$VENV_DIR" ]; then
|
||
echo "✅ venv: $VENV_DIR"
|
||
# faster-whisper version
|
||
if "$VENV_DIR/bin/python" -c "import faster_whisper" 2>/dev/null; then
|
||
FW_VER=$("$VENV_DIR/bin/python" -c "import faster_whisper; print(faster_whisper.__version__)" 2>/dev/null)
|
||
echo "✅ faster-whisper: $FW_VER"
|
||
else
|
||
echo "❌ faster-whisper: not installed (run ./setup.sh)"
|
||
fi
|
||
# CUDA in venv
|
||
CUDA_CHECK=$("$VENV_DIR/bin/python" -c "import torch; print(torch.cuda.is_available())" 2>/dev/null || echo "False")
|
||
if [ "$CUDA_CHECK" = "True" ]; then
|
||
CUDA_DEV=$("$VENV_DIR/bin/python" -c "import torch; print(torch.cuda.get_device_name(0))" 2>/dev/null)
|
||
echo "✅ CUDA in venv: available ($CUDA_DEV)"
|
||
else
|
||
echo "⚠️ CUDA in venv: not available (CPU mode; check PyTorch CUDA install)"
|
||
fi
|
||
# pyannote (timeout 10s to avoid slow CUDA init hanging the check)
|
||
PA_RESULT=$(timeout 10 "$VENV_DIR/bin/python" -c "
|
||
import importlib.util, sys
|
||
spec = importlib.util.find_spec('pyannote.audio')
|
||
if spec is None:
|
||
sys.exit(1)
|
||
# Only read version from metadata, skip full import (avoids 30-60s CUDA load)
|
||
try:
|
||
from importlib.metadata import version
|
||
print(version('pyannote.audio'))
|
||
except Exception:
|
||
print('installed')
|
||
" 2>/dev/null)
|
||
PA_EXIT=$?
|
||
if [ $PA_EXIT -eq 0 ] && [ -n "$PA_RESULT" ]; then
|
||
echo "✅ pyannote.audio: $PA_RESULT (--diarize available)"
|
||
elif [ $PA_EXIT -eq 124 ]; then
|
||
echo "⚠️ pyannote.audio: check timed out (likely installed; run --diarize to verify)"
|
||
else
|
||
echo "ℹ️ pyannote.audio: not installed (--diarize unavailable; run ./setup.sh --diarize)"
|
||
fi
|
||
else
|
||
echo "❌ venv: not found (run ./setup.sh)"
|
||
fi
|
||
|
||
# yt-dlp
|
||
YTDLP_CHECK=""
|
||
if command -v yt-dlp &>/dev/null; then
|
||
YTDLP_CHECK="yt-dlp"
|
||
elif [ -f "$HOME/.local/share/pipx/venvs/yt-dlp/bin/yt-dlp" ]; then
|
||
YTDLP_CHECK="$HOME/.local/share/pipx/venvs/yt-dlp/bin/yt-dlp"
|
||
fi
|
||
if [ -n "$YTDLP_CHECK" ]; then
|
||
YTDLP_VER=$("$YTDLP_CHECK" --version 2>/dev/null)
|
||
echo "✅ yt-dlp: $YTDLP_VER (URL/YouTube input available)"
|
||
else
|
||
echo "ℹ️ yt-dlp: not installed (URL/YouTube input unavailable; pipx install yt-dlp)"
|
||
fi
|
||
|
||
# HuggingFace token
|
||
if [ -f "$HOME/.cache/huggingface/token" ]; then
|
||
echo "✅ HuggingFace token: present"
|
||
else
|
||
echo "ℹ️ HuggingFace token: not found (needed for --diarize; run huggingface-cli login)"
|
||
fi
|
||
|
||
echo ""
|
||
exit 0
|
||
;;
|
||
--help|-h)
|
||
echo "Usage: ./setup.sh [--diarize] [--update] [--check]"
|
||
echo ""
|
||
echo "Options:"
|
||
echo " --diarize Also install pyannote.audio for speaker diarization"
|
||
echo " Requires HuggingFace token at ~/.cache/huggingface/token"
|
||
echo " and model agreement at https://hf.co/pyannote/speaker-diarization-3.1"
|
||
echo " --update Upgrade faster-whisper in the existing venv without full reinstall"
|
||
echo " --check Verify system dependencies (GPU, Python, ffmpeg, venv, yt-dlp)"
|
||
exit 0
|
||
;;
|
||
esac
|
||
done
|
||
|
||
echo "🎙️ Setting up faster-whisper skill..."
|
||
|
||
# Detect OS
|
||
OS="$(uname -s)"
|
||
ARCH="$(uname -m)"
|
||
|
||
case "$OS" in
|
||
Linux*) OS_TYPE="linux" ;;
|
||
Darwin*) OS_TYPE="macos" ;;
|
||
*) OS_TYPE="unknown" ;;
|
||
esac
|
||
|
||
echo "✓ Platform: $OS_TYPE ($ARCH)"
|
||
|
||
# Check for Python 3.10+
|
||
if ! command -v python3 &> /dev/null; then
|
||
echo "❌ Python 3 not found. Please install Python 3.10 or later."
|
||
exit 1
|
||
fi
|
||
|
||
PYTHON_VERSION=$(python3 -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")')
|
||
PYTHON_MAJOR=$(echo "$PYTHON_VERSION" | cut -d. -f1)
|
||
PYTHON_MINOR=$(echo "$PYTHON_VERSION" | cut -d. -f2)
|
||
|
||
if [ "$PYTHON_MAJOR" -lt 3 ] || ([ "$PYTHON_MAJOR" -eq 3 ] && [ "$PYTHON_MINOR" -lt 10 ]); then
|
||
echo "❌ Python 3.10+ required (found $PYTHON_VERSION)"
|
||
exit 1
|
||
fi
|
||
|
||
echo "✓ Python $PYTHON_VERSION"
|
||
|
||
# Check for ffmpeg (required)
|
||
if ! command -v ffmpeg &> /dev/null; then
|
||
echo "❌ ffmpeg not found (required for audio processing)"
|
||
echo ""
|
||
echo "Install ffmpeg:"
|
||
if [ "$OS_TYPE" = "macos" ]; then
|
||
echo " brew install ffmpeg"
|
||
else
|
||
echo " Ubuntu/Debian: sudo apt install ffmpeg"
|
||
echo " Fedora: sudo dnf install ffmpeg"
|
||
echo " Arch: sudo pacman -S ffmpeg"
|
||
fi
|
||
echo ""
|
||
exit 1
|
||
fi
|
||
|
||
echo "✓ ffmpeg found"
|
||
|
||
# Detect GPU/acceleration availability
|
||
HAS_CUDA=false
|
||
HAS_APPLE_SILICON=false
|
||
GPU_NAME=""
|
||
NVIDIA_SMI=""
|
||
|
||
if [ "$OS_TYPE" = "linux" ]; then
|
||
# Check for NVIDIA GPU (Linux/WSL)
|
||
# Try nvidia-smi in PATH first
|
||
if command -v nvidia-smi &> /dev/null; then
|
||
NVIDIA_SMI="nvidia-smi"
|
||
else
|
||
# WSL2: nvidia-smi is in /usr/lib/wsl/lib/ (not in PATH by default)
|
||
if grep -qi microsoft /proc/version 2>/dev/null; then
|
||
for wsl_smi in /usr/lib/wsl/lib/nvidia-smi /usr/lib/wsl/drivers/*/nvidia-smi; do
|
||
if [ -f "$wsl_smi" ]; then
|
||
NVIDIA_SMI="$wsl_smi"
|
||
echo "✓ WSL2 detected"
|
||
break
|
||
fi
|
||
done
|
||
fi
|
||
fi
|
||
|
||
# If we found nvidia-smi, get GPU info
|
||
if [ -n "$NVIDIA_SMI" ]; then
|
||
GPU_NAME=$($NVIDIA_SMI --query-gpu=name --format=csv,noheader 2>/dev/null | head -1)
|
||
if [ -n "$GPU_NAME" ]; then
|
||
HAS_CUDA=true
|
||
fi
|
||
fi
|
||
elif [ "$OS_TYPE" = "macos" ]; then
|
||
if [ "$ARCH" = "arm64" ]; then
|
||
HAS_APPLE_SILICON=true
|
||
GPU_NAME="Apple Silicon"
|
||
echo "✓ Apple Silicon detected"
|
||
fi
|
||
fi
|
||
|
||
if [ "$HAS_CUDA" = true ]; then
|
||
echo "✓ GPU detected: $GPU_NAME"
|
||
fi
|
||
|
||
# Create venv
|
||
if [ -d "$VENV_DIR" ]; then
|
||
echo "✓ Virtual environment exists"
|
||
else
|
||
echo "Creating virtual environment..."
|
||
if command -v uv &> /dev/null; then
|
||
uv venv "$VENV_DIR" --python python3
|
||
else
|
||
python3 -m venv "$VENV_DIR"
|
||
fi
|
||
echo "✓ Virtual environment created"
|
||
fi
|
||
|
||
# Helper: install with uv or pip
|
||
pip_install() {
|
||
if command -v uv &> /dev/null; then
|
||
uv pip install --python "$VENV_DIR/bin/python" "$@"
|
||
else
|
||
"$VENV_DIR/bin/pip" install "$@"
|
||
fi
|
||
}
|
||
|
||
# Install base dependencies
|
||
echo "Installing faster-whisper..."
|
||
if ! command -v uv &> /dev/null; then
|
||
"$VENV_DIR/bin/pip" install --upgrade pip
|
||
fi
|
||
pip_install -r "$SCRIPT_DIR/requirements.txt"
|
||
|
||
# Install PyTorch based on platform
|
||
if [ "$HAS_CUDA" = true ]; then
|
||
echo ""
|
||
echo "🚀 Installing PyTorch with CUDA support..."
|
||
echo " This enables ~10-20x faster transcription on your GPU."
|
||
echo ""
|
||
if command -v uv &> /dev/null; then
|
||
uv pip install --python "$VENV_DIR/bin/python" torch torchaudio --index-url https://download.pytorch.org/whl/cu121
|
||
else
|
||
"$VENV_DIR/bin/pip" install torch torchaudio --index-url https://download.pytorch.org/whl/cu121
|
||
fi
|
||
echo "✓ PyTorch + torchaudio with CUDA installed"
|
||
elif [ "$OS_TYPE" = "macos" ]; then
|
||
echo ""
|
||
echo "🍎 Installing PyTorch for macOS..."
|
||
pip_install torch torchaudio
|
||
echo "✓ PyTorch installed"
|
||
if [ "$HAS_APPLE_SILICON" = true ]; then
|
||
echo "ℹ️ Note: faster-whisper uses CPU on macOS (Apple Silicon is still fast!)"
|
||
fi
|
||
else
|
||
echo ""
|
||
echo "ℹ️ No NVIDIA GPU detected. Using CPU mode."
|
||
echo " If you have a GPU, ensure CUDA drivers are installed."
|
||
fi
|
||
|
||
# Install diarization dependencies (optional)
|
||
if [ "$INSTALL_DIARIZE" = true ]; then
|
||
echo ""
|
||
echo "🔊 Installing speaker diarization (pyannote.audio)..."
|
||
pip_install pyannote.audio
|
||
|
||
# Check for HuggingFace token
|
||
HF_TOKEN_PATH="$HOME/.cache/huggingface/token"
|
||
if [ ! -f "$HF_TOKEN_PATH" ]; then
|
||
echo ""
|
||
echo "⚠️ No HuggingFace token found at $HF_TOKEN_PATH"
|
||
echo " Diarization requires:"
|
||
echo " 1. A HuggingFace account and token (huggingface-cli login)"
|
||
echo " 2. Accept model agreement: https://hf.co/pyannote/speaker-diarization-3.1"
|
||
echo " 3. Accept model agreement: https://hf.co/pyannote/segmentation-3.0"
|
||
else
|
||
echo "✓ HuggingFace token found"
|
||
fi
|
||
echo "✓ pyannote.audio installed"
|
||
fi
|
||
|
||
# Make scripts executable
|
||
chmod +x "$SCRIPT_DIR/scripts/"*
|
||
|
||
echo ""
|
||
echo "✅ Setup complete!"
|
||
echo ""
|
||
if [ "$HAS_CUDA" = true ]; then
|
||
echo "🚀 GPU acceleration enabled — expect ~20x realtime speed"
|
||
elif [ "$HAS_APPLE_SILICON" = true ]; then
|
||
echo "🍎 Apple Silicon — expect ~3-5x realtime speed on CPU"
|
||
else
|
||
echo "💻 CPU mode — transcription will be slower but functional"
|
||
fi
|
||
if [ "$INSTALL_DIARIZE" = true ]; then
|
||
echo "🔊 Speaker diarization enabled (--diarize flag)"
|
||
fi
|
||
echo ""
|
||
echo "Usage:"
|
||
echo " $SCRIPT_DIR/scripts/transcribe audio.mp3"
|
||
echo " $SCRIPT_DIR/scripts/transcribe audio.mp3 --format srt -o subtitles.srt"
|
||
echo " $SCRIPT_DIR/scripts/transcribe audio.mp3 --diarize"
|
||
echo ""
|
||
echo "First run will download the model (~756MB for distil-large-v3.5)."
|