Glossary: Meetings

What is Local transcription?

Turning recorded speech into text on your own computer, so the audio never has to leave it.

Updated Nov 1, 2026By Bruce W.

Local transcription means the speech-to-text model runs on your own computer. The recording is turned into text right there, instead of being uploaded to a company’s servers.

Why people choose it

  • Privacy: the audio stays with you. That matters for client calls, HR, legal and health conversations.
  • No minute limits: your computer doesn’t charge by the hour.
  • Works offline: handy on planes and patchy Wi-Fi.

The trade-offs

  • Hardware: it uses your processor or graphics card, and the model takes disk space.
  • Setup: the model has to be downloaded once.

The longer version is in Local vs cloud transcription.

An example

A recruiter records candidate interviews. A consultant records client calls. A manager records a sensitive one-on-one. With a cloud notetaker, each recording is uploaded to a company’s servers, kept for a while, and covered by that company’s privacy policy. With local transcription, the recording and transcript sit in a folder on the computer that made them. Delete them, and they’re gone.

Local vs cloud at a glance

  • Where the audio goes: local keeps it on your PC. Cloud uploads it.
  • Accuracy: good local models are now close to cloud services for clear speech in common languages. Cloud services can still do better on very noisy audio or rare languages.
  • Speed: cloud is quick on any computer. Local depends on your hardware.
  • Cost: cloud is often priced by minutes or seats. Local has no per-minute cost.
  • Offline: only local works without internet.

Accuracy is usually measured with word error rate.

In Fetch

Fetch transcribes on your computer with two models: a fast one for live notes and a careful one that cleans up after the call. Your recordings and transcripts stay on your PC. Only text goes to Claude or ChatGPT for AI notes, if you connect one. See Choose a speech model.

Questions people ask

Is local transcription less accurate than the cloud?

Not necessarily. Today's open speech models score close to cloud services, especially on a computer with a graphics card.

Does local transcription need a powerful computer?

Not necessarily. Smaller models run on most recent laptops. A faster processor or a graphics card lets you use bigger, more accurate models and keeps the live transcript quick.

Still scratching your head?

Send us a note and a real person will get back to you.

Contact support