Skip to content
← Help

What "local-first" means

If this is your first stop, start here. It explains the one idea that makes Daisy different — local-first — and the one place a cloud service still shows up (summaries, not transcription).

What "local-first" means

Most meeting tools work like this: a bot joins your call, your audio is sent to a company's servers, it's processed there, and your transcripts live in an account you don't control.

Daisy flips that around. It's an app on your computer. Your recordings, transcripts, summaries, and notes are saved as files on your disk, in a folder you choose. By default, your meetings stay on your machine. Recording, transcription, speaker labels, and search all run on-device and keep working offline.

That's "local-first": your stuff lives with you first, and anything online is something you opt into — never the default.

Why it's nice

  • Private — your conversations stay on your device.
  • Yours — the files are right there; back them up, move them, keep them forever.
  • Works offline — on a plane, in a basement, wherever, including through extended periods with no connection.
  • No mystery cloud — you decide if and when anything goes online. We're not interested in your data: aside from the AI summary step you choose, Daisy's only network touches are a periodic license check (at least once every 24 hours), update checks, and activation.

Your transcript is always local

You never have to make a "local or cloud" decision for the transcript you keep. Daisy produces the finished transcript on your machine and figures out who said what on-device — for free, every time. No API key, no per-minute bill. A one-hour meeting finalizes in a few minutes on a typical computer.

The one part that can lean on the cloud is the live caption while you record — and only on lighter laptops:

  • Desktops and Apple Silicon Macs show live captions on-device.
  • Lighter laptops choose: on-device (heavier on the machine), a bring-your-own Deepgram realtime key (lighter), or skip live captions and take the full transcript at the end.

That choice never changes your finished transcript or your speaker labels — those stay local. See Transcription speed and cost for what to expect on your hardware.

The one place a cloud service shows up — summaries

Summaries (the TL;DR, action items, decisions, etc.) are a separate step and they always need an AI model — there's no good fully-offline option for this part yet. You pick how to handle it:

  • Cloud (recommended) — Anthropic, OpenAI, or Groq with your own API key. Best quality; roughly 5¢–50¢ per meeting depending on model and length. The transcript is sent to the provider you choose. Your audio is not.
  • Local AI (advanced) — point Daisy at LM Studio or Ollama running on your own machine or network. Free, runs locally, but only as good as the model you run.
  • Daisy Cloud — a managed option if you'd rather not run a local model or bring your own key. Optional, free for now during a limited public release with no warranties; only the transcript text is proxied (your audio, transcripts, and speaker labels stay local) and no conversations are stored. See What is Daisy Cloud?.
  • Skip — keep transcripts only. The transcript view has a Copy all button — paste it into your own ChatGPT or Claude, ask for a summary, paste the summary back into Daisy.

See AI providers & keys for the full picture.

About the numbers

Any speeds or costs mentioned across the help are rough, order-of-magnitude comparisons. Your real experience depends on your computer, the length of the meeting, and which models you choose. Treat them as relative guidance, not promises.