Repository navigation
Releases
Releases · briankelley/atlas-voice-training
hey_atlas voice models
Pre-trained "hey atlas" openWakeWord model so you can use it without running the GPU training pipeline.
Metrics: 81% accuracy / 62% recall / 55% precision (openWakeWord DNN).
- hey_atlas.onnx - cross-platform (CPU, Apple Silicon); use inference_framework="onnx"
- hey_atlas.tflite - use inference_framework="tflite"