Captions are generated from the speech in an audio clip and arrive as editable text overlays. Fixing a word or restyling every line works the same way it does for titles.
Select the audio clip Click the clip that carries the speech: a voiceover on an audio lane, or the audio that was split from a video clip.
Open Captions In the clip inspector, open the Captions tab. Pick the spoken language (or leave Auto-detect).
Pick a style Standard shows one caption line at a time. Karaoke lights each word up as it is spoken, in a highlight colour you pick. Karaoke uses word-level timing and costs a little more.
Review the price and confirm Click Auto-generate captions. The token cost is shown for confirmation before anything runs, and it covers only the trimmed audio on your timeline.
Wait for the transcription Progress shows in the nav pill, so you can keep editing. Cues land as caption-styled text overlays aligned to the clip.
Edit and restyle Click any caption to fix the wording, retime it, or change its font, outline and box. Captions are ordinary text layers.
Tips
Only the part of the audio on your timeline is transcribed and billed, so trimming first saves tokens.
If no speech is found, nothing is charged.
For multilingual audio, Auto-detect works well on clips longer than a few seconds; pin the language for very short ones.
Karaoke captions keep their word timing through edits. Fixing a typo keeps each word where it was, and a bigger rewrite spreads the spoken window across the new words. Styling changes never touch the sync.