Discussions

Hi HeyGen team,

I'm currently exploring the voice chat integration for my project and came across your built-in voice chat feature which uses your internal LLM. I would like to clarify:

Is it possible to use HeyGen's built-in Speech-to-Text (STT) solution, but instead of routing the transcribed text to HeyGen’s internal LLM, send it to my own custom LLM or backend service for response generation?

In short, I want to:

Use HeyGen's voice activity detection + STT pipeline.

Capture the transcribed text.

Bypass HeyGen’s internal knowledge base/LLM.

Send the transcription to my own LLM and return a custom response for the avatar to speak.

Is this supported in your current SDK/API offering? If not, is there a workaround or recommended integration path?

Thanks!