WhisperForge is a simple web app that helps you collect voice recordings and train a personalized Whisper speech model.
It provides a clean browser-based interface where you can record audio, manage transcripts, and fine-tune a model — all in one place.
- 🎤 Record yourself reading sentences
- 📝 Automatically save audio with matching transcripts
- 📚 Build your own speech dataset
- 🧠 Train a Whisper model using your recordings
- 🔁 Improve your model over time with more data
Start the application and open the following address in your browser:
You will see two main sections inside the app.
- Read sentences directly in your browser
- Record your voice
- Save audio and text together automatically
- Build a clean and organized dataset
- Configure your training options
- Launch the training process
- Fine-tune a model using your recorded data
All recordings and transcripts are stored inside the userdata folder.
This keeps:
- Your sentences
- Your audio files
- Your dataset
Everything is kept separate from the main project files.
WhisperForge is ideal for:
- Creating personalized speech models
- Voice assistant projects
- Research experiments
- Collecting speech data quickly
- Exploring custom speech recognition systems
MIT License