Word timings and forced alignment
When each character and word is said, for captions, karaoke and lip sync.
- Who can do this
- Everyone
- Reading time
- 1 min read
The with-timestamps endpoint returns two alignments. alignment is your text as you wrote it; normalized_alignment is what was actually said. So if you wrote KES 900, alignment times those seven characters across the whole of "nine hundred shillings", and your captions stay in your own words.
Timing audio you already have
POST /v1/forced-alignment takes an audio file and its text, and returns when each word and character is said. It's charged by the second of audio. The whole file is aligned in one language: the one most of the text is in, or the one you name with x-navoras-language.
Still stuck?
A person at Navoras reads every message and replies by email. Name your workspace, and the request ID if it's about one.