4 ms·
We have just updated our public voice detector, now it boasts features like: - Epic metrics (best in class as far as we know); - Much shorter and flexible chun
by snakers41 5y ago
We have just updated our public voice detector, now it boasts features like:
- Epic metrics (best in class as far as we know);
- Much shorter and flexible chunk size (30, 60, 100 ms);
- Radically faster inference, i.e. model forward pass takes ~1ms for a 30 - 100ms chunk;
- Radical simplification - there is literally just one model for everything now;