Private by design
Your voice.
Your browser.
Hold a button or the Shift key and speak. Seda transcribes locally with WebGPU or WASM—no account, no server, and no uploaded audio.
Audio stays local
Open source
WebGPU + WASM
~55 MB model
The first load downloads and caches the pinned English Moonshine Tiny model. Live text is emitted as revisable buffered transcriptions; releasing push-to-talk produces the stable result. One utterance can be up to 30 seconds.
Live transcription
Not loaded
Install speech model
Runs privately in this tab
Transcript
Ready when you are
Your words will appear here.
Hold the button or either Shift key. Release to finish. Press Escape to cancel.