Stop paying for WhisperFlow, Otter.ai, and OCR APIs.
FreeScribe is a 100% free, open-source, offline-first tool for:
- Speech-to-Text — Transcribe audio files (MP3, WAV, M4A, OGG, FLAC, WebM)
- OCR — Extract text from images (PNG, JPG, BMP, TIFF, PDF)
Everything runs locally on your machine. No API keys. No data sent anywhere. No usage limits.
| Service | Monthly Cost | FreeScribe |
|---|---|---|
| WhisperFlow | $15-99/mo | $0 forever |
| Otter.ai | $17-40/mo | $0 forever |
| Google Cloud STT | $1.44/hr | $0 forever |
| Azure OCR | $1/1000 calls | $0 forever |
| Amazon Textract | $1.50/1000 pages | $0 forever |
- Offline — No internet needed after initial setup
- Fast — Uses faster-whisper (5x faster than OpenAI's whisper)
- Multi-language — 99 languages for STT, 100+ for OCR
- CLI + Python API — Use however you want
- GPU optional — Runs great on CPU, faster with GPU
- Docker ready — One command to deploy
pip install freescribegit clone https://github.com/RanaPriyansh/voice-ocr.git
cd free-voice-ocr
pip install -e .docker run -p 5000:5000 freescribe/web# Transcribe audio
freescribe transcribe podcast.mp3
freescribe transcribe interview.wav --language en --model medium
# Extract text from image
freescribe ocr screenshot.png
freescribe ocr document.jpg --language spa
# Process entire folder
freescribe batch /path/to/audio/files/ --output transcripts/
# Launch web interface
freescribe servefrom freescribe import transcribe, ocr
# Speech to text
text = transcribe("audio.mp3", model="base", language="en")
print(text)
# OCR
text = ocr("image.png", language="eng")
print(text)MP3, WAV, M4A, OGG, FLAC, WMA, AAC, WebM, OPUS
PNG, JPG/JPEG, BMP, TIFF, GIF, WebP, PDF
Choose your speed vs quality tradeoff:
| Model | Size | Speed | Quality | Use Case |
|---|---|---|---|---|
| tiny | 75MB | ⚡⚡⚡⚡⚡ | ⭐⭐ | Quick drafts |
| base | 150MB | ⚡⚡⚡⚡ | ⭐⭐⭐ | Recommended |
| small | 500MB | ⚡⚡⚡ | ⭐⭐⭐⭐ | Good balance |
| medium | 1.5GB | ⚡⚡ | ⭐⭐⭐⭐⭐ | High quality |
| large-v3 | 3GB | ⚡ | ⭐⭐⭐⭐⭐⭐ | Best possible |
- OS: Linux, macOS, Windows
- Python: 3.8+
- RAM: 2GB minimum (4GB+ recommended)
- Disk: 500MB-5GB depending on model
- GPU: Optional (NVIDIA CUDA or Apple Silicon)
English, Spanish, French, German, Chinese, Japanese, Korean, Arabic, Hindi, Portuguese, Russian, and 88 more. Auto-detected if not specified.
Install language packs:
# Ubuntu/Debian
sudo apt install tesseract-ocr-spa tesseract-ocr-fra tesseract-ocr-deu
# macOS
brew install tesseract-lang# Dockerfile included
docker build -t freescribe .
docker run freescribe freescribe transcribe audio.mp3We welcome contributions! Areas where help is needed:
- Real-time microphone transcription
- Batch processing improvements
- More language support
- Mobile app
- Browser extension
- GUI desktop app
Q: Is this really free? A: Yes. MIT licensed. Forever free. No catch.
Q: How accurate is it? A: Whisper (the underlying STT model) is near-human accuracy for English. OCR uses Tesseract, which handles clean documents very well.
Q: Does it work offline? A: Yes. After downloading models (happens automatically on first use), no internet needed.
Q: Can I use this commercially? A: Yes. MIT license means use it however you want.
Q: What about private/confidential audio? A: Everything runs locally. Your data never leaves your machine.
If this saves you money, star the repo! It helps others find it.
MIT — Do whatever you want with it.
Built with ❤️ by people who think transcription shouldn't cost money.