Make each cue readable
TTS Lines creates one SRT cue from each generated line. If a line contains too much text for one caption, split it before generating. A natural sentence break often works, but the viewer's reading time matters too.
Check the line text itself. A phonetic spelling used to fix audio may not be the spelling you want viewers to read, so review captions separately.
Export from the final take
Regenerate any line whose words, speaker, or speed changed. Export the full audio and SRT together. The cue for a line lasts as long as its generated clip, and the pause after it moves the next cue forward.
The timing is based on the generated file, not on word alignment. A long line may need to be split into more than one caption in your video editor.
Check after the video edit
Place the narration on the timeline and import the SRT. If you trim, stretch, or move audio, retime the related cues. Watch once with sound and captions before publishing the video.