Silero VAD v5 — ONNX
Voice activity detection for the speech-android SDK.
- ~260K params, <1ms latency per chunk
- Input: 512 samples (32ms @ 16kHz)
- Output: speech probability [0, 1]
Tensors
| Input | Shape | Type |
|---|---|---|
| input | [1, 512] | float32 |
| state | [2, 1, 128] | float32 |
| sr | scalar | int64 |
| Output | ||
| output | [1, 1] | float32 |
| stateN | [2, 1, 128] | float32 |
- Guide: soniqo.audio/guides/vad/android
- Docs: soniqo.audio
- GitHub: soniqo/speech-swift