Does Glimpse work offline?

Yes. Glimpse runs speech-to-text on your device by default, so core dictation works offline on macOS and Windows.

Is Glimpse open source?

Yes. Glimpse is open source under AGPL-3.0, and the desktop app, transcription pipeline, and analytics code are public on GitHub.

Does Glimpse send my audio to the cloud?

Core transcription runs locally by default. Optional cloud speech and cleanup features use provider keys you configure only when you enable them.

Can Glimpse record and transcribe meetings?

Yes. Recording Mode records your microphone and your computer's audio, or only the apps you pick like Zoom or your browser, as separate tracks, with no bot joining the call. A live transcript labels each line by speaker as people talk, using NVIDIA Nemotron 3 Diarization on your computer, for up to 8 speakers. It runs on the same local models you dictate with.

New in 1.3

Dictation that stays on your computer

Press a shortcut, talk, and your words land in whatever app you're using. Record your meetings too, with every speaker labeled. Dictation is free and unlimited, on Mac and Windows.

Download Glimpse

Also on Windows, and open source.

Available on

Your voice, in any app

Glimpse types what you say into whatever you're using, and every word stays on your machine.

Hold your shortcut, talk, and the words land in whatever app you're in.

Good evening

Today

2:34 PM

Hey Sarah, can we push the standup to 3? I have a conflict with the design review.

2:31 PM

Reminder to grab groceries on the way home. Milk, eggs, and that sourdough from the bakery on fifth.

2:28 PM

Just finished reviewing the quarterly report. Revenue is up 12% and retention improved significantly.

How it works

⌥Space

1. Press your shortcut

2. Say what you want to write

Can we push the standup to 3? I have a conflict with the design review.

3. It lands where your cursor is

Why dictation is free

How Glimpse compares Best dictation apps Dictation for Windows

Built-in dictation stops when you pause and gets names wrong. The good alternatives send your voice to a server and charge you every month to use it.

Glimpse runs the speech models on your own computer. Nothing to host means nothing to charge you for, so dictation is free, with no word limit, on every model. It always will be.

A one-time license adds the extras, like meeting recordings and AI cleanup. That's how Glimpse keeps getting better. It's made mostly by one person, and the code is open for anyone to read.

Glimpse for

23:29

pied-piper — zsh

~/pied-piper on main

$ git commit -m "Switch compression to middle-out and fix the Weissman score regression"

[main 4f2a1c9] Switch compression to middle-out and fix the Weissman score regression
1 file changed, 12 insertions(+), 4 deletions(-)

Record Zoom, Meet or Teams with no bot in the call, and follow a live transcript while you talk. NVIDIA Nemotron 3 Diarization labels every line by speaker, on your computer.

Meeting transcription

Where your words go

Glimpse is open source, so you don't have to take our word for any of this.

Privacy page Read the code

Your audio: Transcribed on your computer. Sent nowhere by default.
Your transcripts: Saved on your computer. Sent nowhere by default.
Your API keys: Kept in your settings, on your computer.
Cloud providers: Only if you add one, and only to that provider, with your own key.
Usage stats: Anonymous, and never your words or audio. Off with one switch.