FireRedVAD Stream-VAD — Termux-OS Model Asset
This repository contains a portable FP32 ONNX export of the streaming FireRedVAD model for Termux-OS. It is a raw Model Asset Package, not a service or an executable release.
Contents
| File | Size | Role |
|---|---|---|
model.onnx | 2,287,373 B | streaming DFSMN graph |
cmvn.bin | 640 B | 80 means and 80 inverse standard deviations |
The graph consumes 16 kHz mono features and keeps recurrent state in the consumer. It returns speech probabilities and the next packed state. The package contains no device-specific compiled artifact; a consumer may prepare one in its own local cache.
Provenance
The source material is FireRedTeam/FireRedVAD
at immutable revision 7990aaccc6b7aec1e527743bd30201f2c4a03b8c.
model.onnxis an ONNX export ofStream-VAD/model.pth.tar.cmvn.binis derived fromStream-VAD/cmvn.arkas flat float32 statistics.- The files hosted here are the exact bytes recorded by the Registry.
The local package revision is
1b56350b18bca1243acdc76121acb4c8bbb8d7af.
Runtime boundary
Framework installs and verifies raw bytes. The App/consumer owns model preparation, loading, readiness, recurrent state, and cleanup. This repository does not publish consumer caches or user data.
Links
- Framework: https://github.com/johnson-yo/termux-os-framework
- Speech consumer: https://github.com/johnson-yo/termux_os-service-termux_speech
- Manager: https://github.com/johnson-yo/termux_os-service-hf_model_manager
- Upstream: https://github.com/FireRedTeam/FireRedVAD
See NOTICE.md for provenance and licensing notes.