A fast, accurate desktop app for transcribing audio and video files. Powered by faster-whisper with a clean PyQt6 interface.
Runs on CPU by default for maximum compatibility, with optional GPU acceleration on NVIDIA CUDA systems.
- Accurate Transcription - Powered by faster-whisper (CTranslate2 Whisper implementation)
- Multiple Export Formats -
.txt,.docx,.srt(SubRip),.vtt(WebVTT) - 40+ Languages - Auto-detect or force a specific language
- Drag and Drop - Drop media files directly onto the window
- Timestamps - Optional timestamp insertion in text/docx output
- GPU Acceleration - NVIDIA CUDA with automatic CPU fallback
- Voice Activity Detection - Skip silence to speed up processing (requires
onnxruntime) - Elapsed Time / ETA - Real-time progress with time remaining estimate
- Settings Persistence - Remembers your preferences between sessions
- Same-as-Source Output - Save transcripts next to the original files
- File List Management - See, review, and clear selected files
- Dark Theme - Modern dark UI with styled controls
- Desktop Notifications - Get notified when transcription finishes
- Per-File Error Recovery - Corrupt files are skipped; the batch continues
MP3, MP4, M4A, WAV, FLAC, AAC, OGG, WMA, MOV, WEBM, MKV
- Windows 10/11 (Linux/macOS possible from source)
- Python 3.9-3.12 (if running from source)
- FFmpeg on PATH (required for media decoding)
# Clone and enter directory
git clone https://github.com/HackingPain/TranscribeX.git
cd TranscribeX
# Create and activate virtual environment
python -m venv .venv
# Windows
.\.venv\Scripts\activate
# macOS/Linux
source .venv/bin/activate
# Install dependencies
pip install -r requirements.txt
# Run
python transcribex.py# Windows only
build.bat
# Output: dist\TranscribeX.exetranscribex.py Entry point, FFmpeg check
constants.py Configuration, defaults, TranscribeJob dataclass
worker.py Background transcription thread
gui.py PyQt6 GUI window
requirements.txt Python dependencies
pyinstaller.spec Build configuration
build.bat Windows build script
| Model | Speed | Accuracy | RAM |
|---|---|---|---|
| tiny | Fastest | Lower | ~1 GB |
| base | Fast | Fair | ~1 GB |
| small | Medium | Good | ~2 GB |
| medium | Slow | Better | ~5 GB |
| large-v3 | Slowest | Best | ~10 GB |
Models are downloaded automatically on first use from HuggingFace.
MIT