Offline speech to text
Speech recognition without a server.
Local models, your processor, no network path for your voice. Here is how it works and what it changes.
The model runs locally
Vaitly Voice ships with an on-device English model and can download others. Recognition uses the Mac's neural engine or the PC's GPU. It works in flight mode.
Latency and accuracy
No round-trip means the text keeps up as you speak. Accuracy is tuned for conversational English and improves with your vocabulary.
What still may leave the device
Only optional cloud transforms you configure yourself, clearly labelled. Everything else is local.