Skip to main content
The Audio Transcription plugin adds a Transcribe Files page where you can drop or pick audio files and get clean text back, using the same local or cloud transcription model you’ve configured for dictation.

​
Install

Or install from Settings → Plugins inside the app.

​
Usage

1

Open the plugin page

Once the plugin is enabled, open Audio Transcription from the app sidebar.
2

Drop your files

Drop one or more audio files onto the page, or click to choose from your filesystem.
3

Get your transcript

Each file is transcribed and cleaned up. Copy the result with one click.

​
Supported formats

wav, mp3, m4a, ogg, flac, and webm — plus anything else the browser reports as audio/*. Files are decoded and resampled to 16 kHz mono WAV in the browser before being sent to the server, so any common format works regardless of the provider.

​
Privacy

Audio is sent only to the Freestyle server you’ve configured — your local machine by default. Nothing leaves your device unless you’ve set up a cloud provider.

​
How it works

The plugin’s page talks to the server’s /api/transcribe endpoint through the Freestyle plugin bridge (window.freestyle.api). It ships no model of its own and reuses your configured voice model and cleanup settings. This means:
  • If you’re using a local model (whisper.cpp, MLX), audio stays on your machine.
  • If you’re using a cloud provider (OpenAI, Groq, etc.), audio is sent to that provider — same as dictation.
  • Cleanup intensity and dictionary replacements apply to file transcriptions too.

​
Build from source

This runs pkgroll --minify (server-side plugin code) followed by vite build (UI page).

​
Source

plugins/audio-transcription/ in the monorepo.