⌨
Hotkey-first workflow
Activate Vāk with its configured global hotkey. Recording ends when you release it or the session reaches its time limit.
A focused toolkit
Designed to stay
out of your way.
Vāk keeps the workflow simple: trigger dictation, speak, and return text to the app where you were already working.
◉
On-device recognition
Vāk uses a Whisper model on your computer. Audio is not sent to a cloud speech-recognition service by its transcription pipeline.
↳
Text at the cursor
Vāk uses the clipboard and a paste action to insert text into the active application. Clipboard restoration can be configured.
▤
Recent history
History can be disabled, and entries can be deleted. Diagnostic logs are separate from history.
⚙
English model choices
V1 supports tiny.en, base.en, and small.en Whisper models. Larger models need more storage and processing resources.
◷
Focused sessions
Vāk records on demand while you hold the configured hotkey. It is not designed as an always-on recorder.
A clear note about privacy
Local transcription does not mean no local data is written. Vāk V1 temporarily writes audio to a WAV file during inference, and diagnostic logs can include transcript text even when history is disabled.
Read the full policy →