Offline voice typing for Windows with Music Island

Set up local dictation

1. Download the portable Windows app and open Settings → Dictation.
2. Enable dictation, choose a model that supports your language, and download it. Check the model card and download size before choosing.
3. Select your microphone and set the recording shortcut. Hold, toggle and combined recording modes are available.
4. Focus a text editor, press the shortcut, speak, and finish the recording. The recognized text is inserted into the active app.

What works offline?

Regular speech recognition runs locally. You need network access for the initial model download and software updates, but ordinary recognition does not send your microphone audio to Music Island servers. Settings, recordings and history are stored on your device.

Optional AI text processing is disabled by default. If you configure it and use its separate shortcut, text is sent to the provider you choose. Local recognition and optional provider processing are different operations.

Languages, hardware and insertion

Language support and recognition speed depend on the selected model and your hardware. Choose a model with the language and capabilities you need; do not assume that every model supports every language or streaming mode.

If the target window is closed or cannot accept the result, the overlay offers Copy text. You can also find saved results and audio in History. Deleting the portable EXE does not delete downloaded models or saved data; manage them in System → Data.

Dictation and Better Voice are separate

Dictation recognizes speech and inserts text. Better Voice is an optional beta feature for microphone noise reduction and effects; routing its processed audio into another app requires VB-Cable. You do not need Better Voice to start ordinary dictation.

The dictation engine adapts Handy. Music Island adds it to the same portable desktop app as the music controls; the project is free and open source.