Record
Native audio and video capture (screen, camera, microphone, or system audio) without a browser tab or an upload queue.
An experiment · Open source
Mini Scribe.
A small, local-first tool that records audio and video and turns it into text on your machine, with nothing sent anywhere.
- Status
- Open source experiment
- Home
- miniscribe.app
- Source
- agentic-flow-studios
- Runs
- Fully on device
Press record.
Get words back.
Most recording tools are the front door to somebody's cloud. Mini Scribe is the opposite bet: the machine you already own is fast enough to do the whole job.
Transcribe
Speech to text runs on the machine that made the recording. No API keys, no accounts, no per-minute billing.
Keep
Recordings and transcripts stay as plain files you own, in a folder you chose, readable by anything else you use.
Why build it
this way.
Three constraints that shaped every decision in the project.
Local by default
The recording never leaves the device unless you move it. That is the whole point, not a privacy setting buried three menus deep.
Small on purpose
One job done properly: capture something, get words back. No workspace, no dashboard, no seats to manage.
Open to read
The source is public so the claim is checkable. "It runs locally" should be something you can verify, not something you take on faith.
Use case study
Mini Scribe
in practice.
A working note on where this tool actually earned its place, and where it did not.
In writing. The study will cover:
The situation
Who was recording what, and what it was costing them.
What we tried
The approach, the dead ends, and the version that stuck.
What changed
Observed results: time, cost, and what people actually did differently.
What it means
The transferable part: where local-first tooling earns its place.