How to edit by transcript
- Add your clips and open the Text tab. Under Auto captions, set the spoken language (or leave it on detect).
- Press Transcribe clips in Edit by transcript. The first time, the speech model (about 76 MB) downloads; after that it works offline.
- Mark what should go. Click a word to mark it, Shift-click another to mark everything between. Double-click any word to jump the playhead there and listen.
- Or let it find things: Mark filler words (um, uh, erm…), Mark stutters (“I I think”) and Mark retakes — a sentence you started, stopped and said again; the earlier attempt is marked.
- Press Cut marked words. Every marked stretch is removed and the gaps close. Captions, text, stickers, overlays, voiceovers and music move up with the video. One undo puts it all back.
Features
- Word-level timing — each word has its own start and end, so cuts land between words, taking the pause around a removed word with it.
- Follows your edit — the transcript is grouped by clip and updates when you split, move, trim or speed up clips.
- Find — type a word or phrase to highlight it; Enter jumps to the next match.
- Karaoke highlight — while the video plays, the word being spoken is highlighted.
- Many languages — it uses the same Whisper speech model family as Auto captions.
- Private — the AI runs in your browser; your footage is never uploaded.
Tips for a clean result
- Mark retakes first, then read through: the AI only catches restarts that begin with the same words.
- Combine with Remove silences (Audio tab). The speech model often leaves “um” and “uh” out of the text; the pauses around them are exactly what silence removal catches.
- Don't cut mid-breath — if a cut sounds abrupt, undo and mark one word fewer.
- Add captions afterwards so they're timed to the final edit.
Frequently asked questions
Why are some “ums” missing from the transcript?
Whisper, the speech model, was trained on tidy transcripts and often leaves filler sounds out. When it writes them, Mark filler words finds them; when it doesn't, Remove silences in the Audio tab removes the pauses they cause.
Can I edit the words to fix the text?
The transcript is for cutting — changing a word doesn't change what's said. To fix wording on screen, generate captions and edit them in the Text tab.
Does it work with long videos?
Yes. Only the parts of each file you use on the timeline are transcribed. On a typical computer a 10-minute video takes a few minutes the first time.
Is my video uploaded?
No. The speech model runs on your device, like every AI tool in VDOAPP.