← all projects

whisper online

2026, everything, next.js, react, webassembly

the problem

transcription usually means uploading your audio to someone else's server. i wanted the opposite: drop in a file, get a transcript, and keep every byte on the device.

the approach

whisper running locally via webassembly, with selectable models from tiny to small. token-level editing, correct, merge, or delete individual timed words, plus exports for subtitles, text, and timing data.

the outcome

a focused transcription workspace with guides and zero accounts, uploads, or fees. private by default.

most days it just sits there, quietly turning mumbled voice notes into searchable text.

whisper online transcription page with the upload form and model selector

choose a file and a model; transcription runs on your own device.