Meeting notes without a bot in the call
Pisarik notices when a Teams, Google Meet or Zoom call starts, records it on your computer and, a couple of minutes after it ends, writes the summary: what was decided, who does what and which questions are still open.
Windows 10 (2004+) and 11, 64-bit
Decisions
- The release moves to November 14 to make room for the load test
- The old CSV export stays for one more quarter
Action items
- Anna run the load test by Nov 10
- Me tell customers about the new date
- Marek update the API docs
Open questions
- Do partners need a separate staging environment?
How it works
Nothing to switch on before the meeting, nobody to invite into the call.
The call starts, recording starts
Pisarik lives in the tray. When Teams, Zoom or a Google Meet tab takes the microphone, the icon turns red and recording begins.
The call ends, speech becomes text
Speech is recognized right on your PC: English, Russian and Polish, even mixed within one conversation.
The summary is ready
Decisions, action items with owners and open questions. The transcript, search and a player are on a local recordings page.
Why Pisarik
Made for people who would rather not bring a stranger bot into their calls.
No extra participants
Recording happens on your computer. No “Notetaker” in the participant list, nothing to distract the people you talk to.
You can tell who said what
Your microphone and the other side are recorded on separate channels, so the transcript and the action items know which lines are yours.
Audio stays with you
Recordings and transcripts live in Documents\Pisarik. Recognition runs locally, several times faster with an NVIDIA graphics card.
A summary you can act on
Claude writes the summary in the language of the meeting, with action items and owners. You need your own Claude API key or your team's service.
Good to know
The requirements, and where your data goes.
- System
- Windows 10 version 2004 or later, Windows 11. 64-bit. No administrator rights needed.
- Installation
- The 16 MB installer downloads Python and the recognition components (0.5–1.5 GB). The speech model (1.5–3 GB) downloads at the first meeting.
- What goes online
- Only the transcript text, to the Claude API or your team's service, to get the summary. Audio never leaves your computer.
- Windows warning
- The installer is not signed yet, so SmartScreen may stop it: “More info” → “Run anyway”.
Try it at your next meeting
Install it, add a Claude key, and just get on the call.