MicScribe

Windows-only Electron app for recording a short audio clip and transcribing it via Replicate (model: openai/gpt-4o-transcribe). No live transcription.

Installation

npm install

Environment variable

Use a local .env file for development. The file is ignored by Git and will not be pushed. Copy the template once and fill in your token:

copy .env.example .env

Then edit .env and set your Replicate token:

REPLICATE_API_TOKEN=your-token

If you build/install the app, you can also set the token directly in the UI (Replicate API-Key). The value is stored locally and is not shown after save. Alternatively, set the environment variable at the system/user level.

PowerShell (current session):

$env:REPLICATE_API_TOKEN="your-token"

PowerShell (persist for your user):

setx REPLICATE_API_TOKEN "your-token"

Run

npm start

Build / Installer

npm run make

This creates Windows Squirrel artifacts under out/make/squirrel.windows/.

Usage

• Choose Sprache (German de or English en).
• Select a microphone.
• Click Aufnehmen, speak, then Stop.
• Wait for Transcribing… to finish and copy the result.

Known limitations

• No live transcription (only after recording stops).
• Audio is recorded as WebM (Opus) when supported.
• Requires a working Replicate token and network access.