Open-source credits

Models are downloaded from Hugging Face after you choose to prepare them. Each model retains its own license and usage conditions.

• Supertonic model — OpenRAIL. Default speech model; see its repository for the full terms and upstream attribution.
• Kokoro model and voices — Apache-2.0.
• Meta MMS weights — CC-BY-NC-4.0 (non-commercial).
• Transformers.js, Kokoro.js, Phonemizer.js, Mozilla Readability — Apache-2.0.
• ONNX Runtime, Express and jsdom — MIT.
• Phonemizer includes eSpeak NG — GPL-3.0; source and license are available upstream.

Your passages and recordings stay in your browser. URL import sends the URL to this website’s server and fetches the target page. Model downloads contact Hugging Face and its file hosts. Saved HTML and text files are read on your device.

Generated audio is kept for the current page session. Download recordings you want to keep. Browser settings or storage pressure can remove cached models.