Use case · Privacy-first

For calls the cloud shouldn't keep.

Minuta captures audio locally, sends it through a reviewed cloud AI chain over the internet, and stores transcripts and meeting records in the desktop app. No meeting bot joins the call; review the full data flow on the Trust page.

Before, during and after

Editable summary and autosaved notes in Minuta with synthetic data

Editable summary and autosaved notes in Minuta with synthetic data
Real Minuta app screens with synthetic demonstration data.

When the audio is the asset

Your call has someone who would notice if it were recorded by a third-party processor. The lawyer's client. The patient. The CEO on the other side of the M&A. Or your own security team that flagged cloud uploads. You currently don't transcribe these calls — you take notes by hand, you forget half of them, and the follow-ups are weaker than they should be.

Minuta captures audio locally and routes it through a short, reviewed cloud chain — Deepgram for speech-to-text with OpenAI or Groq as fallback. Internet is required for AI processing, while the transcript file stays in your desktop app's local SQLite database.

Where it earns its keep

Legal / privileged

Client-privileged audio — local capture, reviewed cloud path

You're prepping a deposition or running a privileged client call. Minuta captures system audio on your computer, routes it through a reviewed transcription provider, and stores the transcript in the desktop app. No third-party bot joins. See the full provider list and data flow on our Trust page.

Healthcare / regulated

Patient discussion — controlled payloads, local transcript

Care-coordination call where the audio is sensitive. Minuta keeps the final transcript in your local desktop app and uses a short reviewed cloud chain for AI processing. No third-party meeting bot joins the call.

M&A / exec

The acqui-hire conversation that can't leak

Pre-LOI conversation with the founder you're acquiring. Minuta does not join as a participant, captures system audio locally, routes audio to a reviewed transcription provider, and stores the resulting meeting record in your desktop app.

How it plugs in

Open Minuta a minute before the call. Run the call as you always would — Zoom, Teams, Meet, FaceTime. Minuta captures system audio on the computer and routes it through a reviewed cloud chain; internet is required.

  • Local capture, cloud processing. System audio is captured on your computer and sent to Deepgram for transcription, with OpenAI or Groq as fallback.
  • No bot in the call — Zoom, Teams, Meet, FaceTime, native phone bridge, it doesn't matter. The other party doesn't see a third-party participant.
  • Transcript stored locally. The transcript file lives in the desktop app's local SQLite database on your machine. Encrypt your disk, back up to your own NAS, delete after your retention period.
  • Pro features such as copilot answer drafts and summaries also use the reviewed cloud AI path described on the Trust page.

Why a reviewed cloud chain is a different category

  • Minuta documents the reviewed cloud chain — Deepgram, OpenAI, and Groq — and what each provider receives on the Trust page.
  • No bot in the call is a hard requirement, not a feature. The presence of a third-party recorder in a privileged conversation is a legal liability.
  • Narrow payloads. Minuta sends audio, gets transcript text back. It does not send full meeting recordings to a cloud storage bucket.
  • See every provider and what they see in our Trust page — the full written data-flow, not a marketing summary.

Other situations Minuta is built for

More ways professionals stay sharp on live calls across English, Portuguese, and Spanish.

Other honest comparisons

Pick the meeting tool that actually matches your call.

Try it on your next sensitive call

Try every feature for 7 days, with 120 minutes total. After that, Free includes 60 shared transcription and translation minutes per month; copilot answers require Pro. See current plans.

Start free trialDownload the desktop app

Keep exploring