Capture
Low-latency recognition turns a live voice into readable text while you speak.
PUBLIC SERVICE INFORMATION
Comma’s online service keeps context across pauses, so a thought can arrive in pieces and still become one clear record.

Every fragment belongs to the same thought.
THE SERVICE SHAPE
Hold-to-speak starts the capture, release-to-pause protects the fragment, and continue-later keeps the record open until the idea is complete.
Low-latency recognition turns a live voice into readable text while you speak.
Each pause is a deliberate boundary. The server keeps the record ready for the next segment.
New segments join the same history record, preserving a coherent thread.
PRIVACY ARCHITECTURE
On-device recognition keeps audio on the phone. Online recognition sends only what is needed through Comma and stores encrypted transcript text for up to 30 days before automatic cleanup.
Read the privacy policy →RESOURCES