Speech marks
Know exactly when each word is spoken for captions and highlighting.
- Word-level timestamps with synthesis
- Powers synced captions (WebVTT, SRT)
- Enables read-along word highlighting
- Returned alongside audio
Definition
Speech marks are word-level timestamps returned with the synthesized audio. They tell you the exact moment each word is spoken. That timing is what powers synced captions and read-along word highlighting, features that would otherwise require a separate transcription and alignment pass.
Synced captions
Accessibility rules often require captions. With speech marks, you map each word’s timestamp into a caption format such as WebVTT or SRT and get accurate synced captions straight from the synthesis, no manual transcription and no drift.
Word highlighting
A read-along experience highlights each word as it is spoken. Speech marks provide the per-word timing that drives that highlight, which helps readers with dyslexia and low vision follow the text.
How to use
Request speech marks with the synthesis call and they return alongside the audio. See the speech marks guide for the response shape and mapping examples.
Frequently asked questions
What are speech marks in text-to-speech?
How do I make synced captions?
Can I highlight words as they are read?
Start building
Know exactly when each word is spoken for captions and highlighting.