How to turn a recording transcript into captions

Captions that do not match the picture are a second mistake. That usually happens when the caption file was built from an older take, or from a title, or from a model that never heard the audio.

Caption the file you will publish

If you had a script, run Script vs delivery and retake the misses that matter. If the words are right and the file is slow, Dead-air and filler cut can build a shorter copy. Then caption that copy. Do not reuse an SRT from the uncut take — the times will slip.

Download SRT or VTT from the transcript

Upload the locked file on the home page. Wait for the timestamped transcript. A registered account with a verified email can download SRT and VTT. Open that file in a caption editor. Fix names, slang, and “uh” you want gone. Do not write a funnier line than the audio has.

Watch the first minute with captions on before you export. If a line sits on the wrong shot, the times are wrong — do not “fix” it by rewriting the words.

Description timestamps

Chapters on the results page can become description timestamps only if they still match the published cut. After a dead-air cut, regenerate or check those times. A chapter list from the long take will send people to the wrong second.

This is not a tool that places captions on the frame. You will still hand-fix. Speech-to-text misses names. That is expected.