## Lecture Video
### Updates & Improvements
* Voice feedback generation is more robust with an improved bufferer for
streamed text chunker.
* ElevenLabs chunk length schedule mirrors the default setup in the SDK.
* The WebSocket connection task to ElevenLabs gets initiated earlier in
the response generation cycle to optimize latency.

### Resolved Issues
* Fixed: Lecture Video voice feedback may stop early or skip later parts
of an assistant response.
* Fixed: Lecture Video voice feedback may sound rushed or unclear, with
some words blending together.