New release! Flux word-level timestamps are here. 🎙️ Every word.

Deepgram's Flux streaming transcripts gained a start and end time for each word. That lets an application locate individual words within the audio rather than work only with the transcript text. The June 30 release notes document the added fields in the words array returned by the /v2/listen endpoint.

The notes call the change additive, but say a client library that rejects unrecognized response fields should be updated to a current SDK version; for the Python SDK they name 6.0.1 or later, or 5.3.4 on the 5.x line. The release therefore adds finer timing information while creating a compatibility check for applications with strict response validation.