thanhng8/supertonic-tool

supertonic-tool is a small toolkit built on top of Supertone's Supertonic, a text to speech engine, that turns it into a local server you can talk to over WebSocket along with a browser based interface for using it. It is a companion project, not part of the original Supertonic repository, and it stays in sync with that upstream project through a helper script. The server automatically detects the best way to run the speech model on your machine, checking for CUDA on NVIDIA graphics cards, DirectML on Windows, or CoreML on Mac, and falling back to plain CPU processing if none of those are available or if an accelerator fails to start. When several text to speech requests arrive close together, the server groups them into a single batch and processes them with one call to the model instead of one call each, which speeds things up noticeably when converting many lines at once. The included web page lets you build a queue of dialogue lines, each with its own voice and language, which is useful for multi speaker conversations or narration with different characters. You can paste in multiple lines at once or import them from a spreadsheet, CSV, or text file, and there is a shorthand format for setting the voice and language inline. Once converted, each line gets its own audio player and download button, and the whole batch can be downloaded together as a single zip file of audio files. To get started, you first set up the original Supertonic project, then run one of the included launcher scripts, which install any missing dependencies and start the server. A Windows batch file and a shell script are both provided, and there are example client code snippets for browsers, a Chrome extension, Node.js, and Python for anyone who wants to connect their own program instead of using the included web page. Python 3.10 or newer and the uv package manager are recommended for the smoothest setup.