Upload a raw talking-head recording, get a finished short back: the pauses and filler words cut out, word-by-word captions burned in. Two things are different. The cut points come from a loudness measurement of your original audio, not from speech-recognition timestamps - those were off by up to 0.9s in testing, which is why some tools clip the end off a word. And rendering runs in your browser via WebCodecs, so for the cut the video never leaves your machine. First three videos free.
Hi Product Hunt 👋
I run a small software agency and make videos to get clients. The part I kept putting off was always the same: cutting out every pause and "uhm", then adding captions because most people watch without sound. For a ten minute recording that is around fifty tiny cuts in a timeline, and not one of them is an interesting decision.
So I built Caption & Cut to do that pass for me. You upload a raw recording and get it back with the silences removed and word-by-word captions burned in. Optionally it also removes filler words and the takes you started over.
Two things I did differently:
1. The cut points come from a loudness measurement of your original audio, not from the speech-recognition timestamps. Those look precise but were off by up to 0.9s in my tests — on one clip the word "Captions" ends at 23.30s according to the transcript and is actually spoken until 24.20. Cut on the transcript and you slice the word in half at full volume. I spent a week stacking correction rules on those timestamps, five of them, each with a real reason and each producing the next class of defect. Deleting all five and measuring the audio instead was 460 lines less code and measurably better output.
2. Before anything renders you see the transcript with every planned cut struck through, and one click puts a word or a pause back. A tool that silently drops a sentence that mattered has ruined the video, and you only hear it afterwards.
Rendering runs in your browser via WebCodecs, so for the cut itself the video does not have to go anywhere. The audio track does go to a hosted Whisper for the transcript — that part is not local and I would rather say so than imply otherwise.
What it is not: it is for talking-head material, not music or multi-camera edits, and it does not tell speakers apart.
Three videos are free, nothing to install. I would really like your feedback, especially on where it still cuts badly on your own footage.