# oneread

> oneread is a private text-to-speech library you run yourself. Paste text or drop in a Word, PDF, Markdown or slides file, pick a voice, and get back a WAV with an SRT timed to it. Speech is synthesised on your own machine by the Supertonic 3 model running under ONNX, so there is no API key to hold and no per-character bill at the end of the month.

Self-hosted, open source, no per-character billing. Speech is produced on the
host by Supertonic 3 running under ONNX Runtime; the model ships inside the
Docker image. Every reading comes with an SRT whose cue times are derived from
the audio samples actually written, so subtitles and audio match exactly.

## Docs

- [About and FAQ](https://oneread.amitj.me/about): what it does, how to run it,
  the answers to the usual questions.
- [Full text](https://oneread.amitj.me/llms-full.txt): every feature and every
  FAQ answer in one file.
- [Source](https://github.com/oneamitj/oneread): code, issues, installation.
- [Speech model](https://github.com/supertone-inc/supertonic): Supertonic 3.

## Notes

- The app itself, at `/`, is behind a sign-in and holds private per-user
  libraries. There is nothing to crawl there.
- Not an API service: it is software you run. There is no hosted endpoint to
  call and no key to obtain.