Launching today
VocaScript
Transcribe hours-long recordings, reliable to the end
12 followers
Transcribe hours-long recordings, reliable to the end
12 followers
VocaScript transcribes hours-long recordings without requiring you to split them. You get timestamped, speaker-labeled text in 100+ languages, even when speakers switch languages. Each transcript is checked against the audio to auto-fix any issues it can, and a quality report walks you through the rest. Need a passage redone? Retranscribe it surgically. Upload, paste a link, or transcribe as you browse with the extension and read along with captions on the page. No signup needed to try it.
Hey Product Hunt, I'm Diae and I made VocaScript.
It started simple. Reading is faster than listening, so I wanted to read and search through recordings instead of having to sit there and listen until the end. I needed something that could take anything from WhatsApp voice notes to long podcasts, lectures or even 10 hour streams, some of them in several languages with 5 or more people talking.
The tools I tried couldn't. They gave up on the long ones, or gave back text I couldn't trust. How hard could it be? I'd just build my own in a week or two (famous last words, I know). That was 8 or 9 months ago.
Since then it's been nothing but hurdles. Long recordings would skip whole chunks of speech or make up words over silence. Timestamps would drift or just be made up. Speaker labels would split one person into three or put two people under one.
What I'm launching today catches and fixes most of that before you even see the transcript. It's not perfect, but I think it does this better than anything else out there.
But we all know a transcript being ready doesn't mean you're done. You still have to go through it yourself and find what's wrong, and that's the slow part. A lot of the work went into making that part easy.
The quality report lists every spot that still looks off and takes you straight to each one. You tick what you want fixed and it does it: made-up and repeated lines get removed, timestamps get corrected. If a stretch of speech is missing, it transcribes just that stretch.
Speakers work the same way. The report tells you when two labels sound like the same person, and it can merge them by voice for you. Or you open the speaker list, listen to a short clip of each one, and rename or merge them yourself.
You don't have to bring it a file either. With the extension you can transcribe as you browse: one click on whatever you're watching or listening to, and the captions show up right on the page so you can read along.
There's a lot more around the transcript too: translation, summaries, proofreading, chat with your transcript, word timings, an editor, search, and exports to TXT, SRT, VTT and JSON.
You can try it with no signup (5 minutes a file). A free account (no card needed) does 25 minutes per run, and you can run it again on the same file to pick up where it stopped, up to 2 hours a month. Paid plans start at $10 a month, take recordings of 4, 8 or 16 hours, and raise the limits across the board.
If you have a long recording that other tools gave up on, try it and tell me where it goes wrong, that's the feedback I need most. I'll be around all day.