Read It, Dude bundled Supertonic 2 voices

Model: supertone-oss-archive/supertonic-2
Pinned revision: 7edbcd0bba8cd579100a353a9e0b2b49cad7c875
Model and voice styles: BigScience Open RAIL-M, see LICENSE.txt
Voice presets: F2 female and M2 male; multilingual presets, no formally documented US accent
Code: Supertone browser example, MIT, copyright 2025 Supertone Inc.
Runtime: ONNX Runtime Web 1.30.0, MIT

Modified by Glitchmodo on 2026-10-05: four existing cross-attention tensors exposed as additional outputs for word alignment; neural operations and learned weights unchanged
The modified model remains subject to the original model license and use restrictions
helper.mjs changes: local runtime imports, correct Supertonic 2 language list, capture of the final two inference steps' cross-attention outputs, and a verified local-cache model loader
Speech corrections and alignment code: Glitchmodo
No recognition model or remote speech service is used
Firefox downloads the exact full-precision model files after installation; the runtime, configuration and voice presets are included with the extension