Discussions

I’m trying to remove the green background from transparent avatars in LiveAvatar.
I’ve tried multiple methods using canvas elements and RGB manipulation, but the quality drops and it doesn’t look as clean as the preview shown on HeyGen.
Any suggestions on how to achieve a true transparent background during streaming?

Hello!
I have two voice clones under ‘My Voice’. I can access the delete menu item, but deleting does not work – nothing happens. The window displayed freezes.
Best regards

Hello,

The request consistently returns a 401 Unauthorized error (HTTP 401).

Hello!

having an issue connecting conversational/interactive agent here is what Replit is saying: are there any workarounds Apparently this wont work with Replit, I am building a prototype here is what it is saying.Same CSP errors in production too. This confirms it's not a Replit development restriction - the HeyGen SDK fundamentally doesn't work in environments with strict CSP policies.
a5b7358c520b4e8b89d534f7c2d12867 • 13m
Even though I've disabled CSP in our code, Replit's infrastructure is injecting its own CSP headers that we can't override. Notice the error is coming from __replco/static/devtools/injected.js - that's Replit's own tooling enforcing the policy.
Here are your options:
Option 1: Use Simli Instead (Already Integrated ✅)
You already have Simli AI working, which provides similar photorealistic avatars. It's already configured and ready to go - would you like me to create an AI Demo V3 using Simli?
Option 2: Deploy to Production
The CSP restrictions might not apply in Replit's production deployment. Would you like me to test this by publishing the app?
Option 3: Contact HeyGen Support
This might be a known issue with their SDK in restrictive environments. They may have a workaround or alternative integration method.

Used a template. Tried using brand glossary, that didn't work. Did preprocessing of a script using n8n functions. tried every combination of SaaS (sass, sas, s-a-a-s) and it keeps reading the lettings instead reading it as a word. Is there anyway to fix this?

I'm trying to use offset to position the avatar using the API. But as soon as I specify the offset object I get a video without the avatar visible back.

I need to confirm which voice I used in avatar videos. How do I do this?