TL;DR — A free, open-source Mac app that writes down what you hear in a class, a video or a meeting, live, and
answers your questions about it in a floating panel. Speech can be transcribed entirely on your Mac with Apple
Speech; answers come from the AI you choose: a ChatGPT sign-in, an API key, or a local model. For Apple Silicon Macs
on macOS 14 or newer. Download it, install it with Homebrew, build it from source, or let your coding agent install
it for you.
https://github.com/user-attachments/assets/ec2e9893-6e2d-42a9-a7cd-04db263d3fdc
Contents
What it does
- Live transcript. While a class, a meeting or a video plays, the words appear as people talk. They come from Apple
Speech on the Mac, a local Whisper server, or a cloud transcription service, whichever you configure. Names and
terms you add under Vocabulary help local Whisper and cloud transcription spell them right.
- Ask about what was said. A floating panel sits over any app. Ask a question and the answer streams back, grounded
in the transcript. A screenshot or the text you have selected can go with the question.
- Assistants for the moment. English, Math, Programming, a Note Taker for meetings and a General assistant come
with the app, and you can make your own.
- Session history. Reopen transcripts and answers, generate a summary, or save a Markdown note.
- Answer providers. Sign in with ChatGPT, use an API key for OpenAI, OpenRouter or Gemini, connect a custom
OpenAI-compatible endpoint, or download and run a local model.
Get it
Download the latest DMG
from GitHub Releases, for Apple Silicon Macs on
macOS 14 or newer, and drag LearnTranscriber to Applications.
The app is not notarized by Apple, so macOS blocks its first launch: open System Settings → Privacy & Security,
scroll to Security and choose Open Anyway next to LearnTranscriber, then confirm with your password
(Apple's instructions).
When a new version is out, the Home banner downloads it and keeps it only if its SHA-256 matches the one the release
publishes; choose Open installer and drag LearnTranscriber over the previous copy. Your sessions and settings stay.
Or install it with Homebrew, which handles updates too:
brew tap learntranscriber/tap
brew trust learntranscriber/tap
brew install --cask learntranscriber
brew trust is Homebrew's one-time consent for a tap it does not maintain. The first launch still needs
Open Anyway; later versions come with brew upgrade --cask learntranscriber.
You can also build it from source or let your coding agent install
it.
Install with your agent
If you use a coding agent such as Claude Code or Codex, send it the link to this repository and ask it to install the
app, or paste this:
Install LearnTranscriber on this Mac from https://github.com/DeVFirmino/learntranscriber.
Clone the repository and follow the "Build from source" section of its README: check that Xcode with the macOS 26 SDK
and Node.js 24 or newer are installed, run `npm ci` and `npm ci --prefix ui`, build the app with `npm run pack`, copy
dist/mac-arm64/LearnTranscriber.app to /Applications, and open it. If macOS blocks it, tell me so I can choose
Open Anyway in System Settings → Privacy & Security.
The agent can do everything but the first launch: there, macOS asks you for each permission and, for a build without a
Developer ID, for Open Anyway.
First run
After the welcome screen, guided setup covers four areas. Skip for now closes setup, and Continue setup on
Home reopens it.
- Choose the AI that answers: sign in with ChatGPT, paste an API key, or download a local model. The step checks
the connection and picks a model before moving on.
- Set up listening: test your microphone with a live level and choose an audio source and transcription engine.
Setup defaults to computer audio alongside the microphone and on-device Apple Speech.
- Grant permissions: setup guides you through the permissions required for your choices, one at a time.
- Try a session: the floating bar and panel guide you through starting capture, getting a transcript and asking
a question. If setup is incomplete, you can play a sample conversation.
How it works
The app is an Electron main process written in Node, a React renderer, and a set of small Swift helpers.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
