Docs
Voice memos & meetings
The same local Gemma 4 model has a native audio encoder — so ScreenMind records and transcribes voice without bolting on Whisper or any cloud speech API.
Voice memos
Hold the voice hotkey (Ctrl+Shift+V by default), speak, and release. Gemma transcribes it locally and captures a screenshot alongside, so the note is tied to whatever you were looking at.
Meetings — auto-detected
1
Detect
When a meeting app (Zoom, Teams, Meet, Discord) comes to the foreground and voice is detected, recording starts on its own — nothing to click.
2
Transcribe
Audio is transcribed in ~15-second chunks by Gemma's audio encoder as the meeting runs.
3
Summarize
On stop, a structured summary is produced — topics, decisions and action items. Long meetings use map-reduce over the transcript.
Recording stops on sustained silence, when the meeting app closes, or a safety timeout — so it never records forever. Transcripts are searchable and show up in chat and MCP.
Fun fact: because it's one model, meeting audio, screen vision and chat all share the same 4 GB GPU — no separate speech model, no extra downloads.
↯ back to the brain