OpenWebTTS: Local Text-to-Speech Web UI

OpenWebTTS is the open-source, privacy-first alternative to Speechify and ElevenLabs. Run it locally, use any TTS engine, and read PDFs, Epubs and other documents without subscriptions or tracking.

Better than paid alternatives

• Clean Interface: Straight to the point and no ads, simple by design, powerful if needed.
• Accessible Design: TTS readers help with many disabilities such as dyslexia and ADHD. We welcome neurodivergence and include accessibility options everywhere possible.
• Multiple Engine Support: Options for any type of hardware, and even cloud options if wanted.
• Voice cloning*: With a simple 10 second wav file you can clone any voice to read for you!
• Import anything: Most document types are supported, and URLs too!
• Automatically skip headers and footers*: Premium feature no more!
• Automatic OCR*: If your PDF doesn't have text, we can make some for you.
• Offline first*: No connection neeeded.
• Self-hostable: Take control of your data, with no feature locked away.

Features marked with an * are paid on other platforms!

Running

See BUILD.md for detailed instructions. If you know what you're doing: clone the repo, install Python dependencies with a venv and build with npm.

Using TTS models

Piper

1. Use the integrated model downloader (recommended)

Or

1. Download a Piper voice model from the official repository.
2. Place the files inside models/piper/. For example: models/piper/en_US-lessac-medium.onnx and models/piper/en_US-lessac-medium.onnx.json.

Kokoro

1. Use the integrated model downloader (recommended)

Or

1. Download a model from the official repository.
2. Place the file inside models/kokoro/. For example: models/kokoro/af_heart.pt

Coqui

Coqui downloads itself automatically with Python. Currently we only support XTTS2, with YourTTS coming soon. We don't plan on supporting every Coqui version, as it will be mostly used for voice cloning since other models have since surpassed it in regular TTS.

1. Place the audio files for voice cloning inside models/coqui/. For example: models/coqui/my-voice.wav.

Chatterbox (WIP)

Chatterbox will use the same audio files for voice cloning as Coqui, so the proccess is the same.