Willow Scribe: speak the gist, get a finished message

Willow, the voice dictation tool that processes speech in under 200 milliseconds, launched its Scribe feature on Product Hunt on May 26, 2026. Where the base product transcribes what you say verbatim, Scribe takes rough spoken input — the way you’d explain something to a friend, mid-thought and unedited — and produces a finished message written in your voice.

The feature runs inside the applications where most writing already happens: Gmail, Slack, iMessage, and Google Docs. Users can also highlight any text on screen and describe how to change it; the selection is rewritten in place without switching applications. The model is trained to maintain the writer’s register rather than smooth everything into a uniform, neutral tone.

For writers and communicators who think verbally before putting ideas to text, Scribe removes the transcription-to-editing step without requiring a change in tooling. The practical constraint is the same as for any dictation tool: it requires speaking where you might otherwise type in silence, which limits use in shared or quiet workspaces.