How do you create captions from generated narration?

How do you create captions from generated narration?

A script contains intended words. Captions need words and timing that match the actual audio, including meaningful sounds where relevant.

Choose a timing source

Use reliable speech-generation timing data when available, or transcribe and align the final narration. Compare the resulting text with the approved script and listen for differences. Correct names and punctuation without inventing speech that is absent from the audio.

Review as timed media

Check cue start and end times during playback. Include useful speaker identification or non-speech information when needed to understand the content. Inspect the beginning, transitions, and ending rather than only a single sentence. If the audio changes, regenerate or realign the captions. Treat automatic caption output as a draft that needs accuracy review before it becomes part of the final deliverable.

Further reading: W3C guidance on captions.