Discussions

Hi HeyGen Team,

Hope you're well.

We’re currently testing interactive avatars for high-scale use (approx. 100,000 videos over the next 6–8 months), but we’re noticing a critical gap in the experience.

The avatars tend to cut off users mid-sentence or respond too quickly without allowing natural conversation flow. We’ve explored your platform settings but haven’t found an option to configure pause duration or voice activity detection (VAD) behavior—something we’ve seen well-supported on platforms like OpenAI (ref).

We’d like to request:

A way to increase the pause sensitivity before the avatar responds

Ensuring the avatar waits for the user to finish speaking

A slower, more natural speech pace for the avatar to mirror real conversations

Is this something your team can configure for us, or is there an upcoming release that will address this?

We’re excited about the potential, but this adjustment is key to moving forward with volume.