BMONI agent voice bundle
The speech models the BMONI app downloads for the on-device agent's voice chat. Each zip holds two models, unchanged from their upstream sherpa-onnx releases:
| Folder | Model | Upstream release | Licence |
|---|---|---|---|
stt/ |
Moonshine Tiny (English, quantized) | sherpa-onnx-moonshine-tiny-en-quantized-2026-02-27 |
MIT (licenses/moonshine_mit.txt) |
tts/ |
Supertonic TTS (int8) | sherpa-onnx-supertonic-tts-int8-2026-03-06 |
Code MIT (licenses/supertonic_mit.txt), model OpenRAIL-M (licenses/supertonic_openrail_m.txt) |
The Supertonic model is distributed under OpenRAIL-M, including its use restrictions, which apply to anyone who uses or redistributes it.
Files
| File | SHA-256 | Bytes |
|---|---|---|
agent-speech-2026-11-09-v1.zip |
1811730dead5b85fc70cdde19a12b41dd0b658bcc48a2442838d92302ad2180c |
110393673 |
Built with scripts/build_speech_bundle.sh in the BMONI mobile repository, which
checks both upstream archives against pinned SHA-256 hashes. Each version gets a new
file name; existing files are never replaced.