| license: mit | |
| tags: | |
| - speech | |
| - vad | |
| - coreml | |
| - voice-activity-detection | |
| # FireRedVAD — CoreML | |
| CoreML conversion of FireRedVAD for voice activity detection on Apple Neural Engine. Supports 100+ languages with high accuracy (99.1% F1 on FLEURS). | |
| ## Files | |
| | File | Description | | |
| |------|-------------| | |
| | `fireredvad.mlmodelc` | VAD model (CoreML compiled) | | |
| | `fireredvad.mlpackage` | VAD model (CoreML package) | | |
| | `cmvn.json` | Feature normalization statistics | | |
| | `config.json` | Model configuration | | |
| ## Usage | |
| Used by [speech-swift](https://github.com/soniqo/speech-swift) `SpeechVAD` module: | |
| ```swift | |
| let vad = try await FireRedVADModel.fromPretrained() | |
| let segments = vad.detectSpeech(audio: samples, sampleRate: 16000) | |
| ``` | |
| --- | |
| - **Guide**: [soniqo.audio/guides/vad](https://soniqo.audio/guides/vad) | |
| - **Docs**: [soniqo.audio](https://soniqo.audio) | |
| - **GitHub**: [soniqo/speech-swift](https://github.com/soniqo/speech-swift) | |