Transcription
Who said what, in which language — transcribed on your computer
No Prod AI answers the three questions about a recording’s speech without sending it anywhere: what was said and when, in what language, and by whom. With a camera on each speaker, it adds the expression on their face.
Free for 14 days, no card needed.
How it goes
- 1
What was said, and when
Whisper, running on your machine, writes every word with its timing.
- 2
In what language
Decided by voting across the whole recording, then held — so a French episode is written in French, not translated into English.
- 3
By whom
Speaker models tell voices apart and number them in the order they first talk.
- 4
With what expression
From each speaker’s own camera, the expression on their face while they talk, and the most expressive frame of each line.
Never uploaded
The models are fetched once, on the first launch, and run on your computer. The recording never leaves it.
Searchable
Search a transcript by speaker:, lang: or emotion:, or ask the assistant to find where something was said across your library.
It feeds everything else
The same transcript cuts the episode, finds the moments, writes the chapters and the captions.
Questions
Does transcription need an internet connection?
Only once, on the first launch, to fetch the models for your machine. After that it transcribes with no connection at all.
Are speakers named?
They are numbered — Speaker 1, Speaker 2 — in the order they first talk. Naming the same voice across a series is not built yet.
Can it translate?
No, deliberately. It writes the language that was spoken. Translating by accident is the mistake it is built to avoid.