Speech recognition
Turn speech into text on your computer or in the cloud.
Recognised text lets Kaddro find retakes by repeated phrases, cut whole attempts near markers, and power AI suggestions. It's also handy for searching: click a word and the video jumps there.
On your computer
- Choose the language or leave it on auto-detect. About 99 languages are supported.
- Choose a model. The first time it has to be downloaded — 550 MB to 3.1 GB. The fast model suits most recordings; the full one is slower but surer on hard passages.
- Start recognition. Progress and the time left show on the timeline; you can cancel.
The speed test shows how long recognition will take on your computer. The text is remembered per recording: open the same recording in another project and it needn't be recognised again.
In the cloud (Pro)
With a Pro subscription you can recognise in the cloud — no model to download, and much faster on slower computers. The sound is sent in pieces to our recognition service and not kept. The quota is 20 hours of audio a month; what's left shows alongside.
If the speed test showed your computer would need more than 10 minutes per hour of speech, Kaddro offers the cloud by default.
Working with the text
- Words recognition is unsure of are underlined.
- Search the text: Enter goes to the next match.
- Copy the text with or without timecodes.
- Re-recognise a single sentence — with another model if needed.
- Change the text size, or clear the whole project's text.