Transcript to Clean Notes
Turn short English audio or video into an editable transcript, then make tidy, source-grounded notes. Everything you provide is processed in this browser tab.
Choose a source
Upload a recording, or paste a transcript you already have.
Up to 15 minutes and 250 MB. Common formats work when your browser can play their audio. Video may need to play through once for local capture.
First, the browser extracts sound. Then a small English speech model converts it to text. GPU acceleration is used when available; CPU processing can take longer.
For video that cannot be decoded directly, press play in the preview when prompted. Capture runs at normal playback speed.
Review the transcript
Edit names, numbers, and anything the speech model missed before refining.
Refine into notes
Choose a format. The local refiner removes common speech fillers and keeps statements tied to the transcript.
What to expect
The first transcription downloads an English Whisper model and its browser runtime. Transcription speed depends on your device. The notes pass is extractive: it organizes and cleans what the transcript says, without adding facts or identifying speakers. Check the output before sharing it.
Speech recognition uses Transformers.js and Whisper Tiny English. Neither audio nor text is sent to a transcription API.