I’m running into two problems when using Convai Studio:
Chatting with the avatar produces no responses.
I can type messages in the chat window, but the avatar never replies.
It looks like the request is sent, but there’s no text or audio response back.
I checked my setup — the avatar is published, and I’m testing it directly in Convai Studio.
Avatar does not work with any external HTTP data.
I tried connecting the avatar to an external API (to feed in live data), but it doesn’t seem to support external HTTP calls.
The avatar just ignores it, and I don’t see any data flowing in.
I’m not sure if I’m missing a configuration step, or if Convai avatars can’t natively call external APIs from Studio.
What I’ve Tried
Verified my API key is valid.
Checked the Response Source in the avatar’s settings (Convai LLM vs External API).
Looked in the browser console for errors (I saw requests going out, but no response payloads).
Tested with both a default Convai LLM setup and a custom endpoint.
My Questions
Why is the avatar not replying in Studio chat — is there another setting I need to enable for conversation responses?
Does Convai Studio support external HTTP API calls directly, or do I need to set up a middleware service to handle that data and then connect it through Convai’s external API integration?
Any guidance from the community or Convai team would be super helpful!
I’ve had the same issue, with multiple collaborators reporting non-response from my character when they follow the published link. Here’s a screen capture of the problem as of today, 10/20/2025. Any guidance toward a solve would be greatly appreciated.
Please log out and back in at x.convai.com. If the issue persists, try an incognito/private window, and then test with a different browser. If it still continues, open the browser DevTools with F12 and share the console logs.
Your note reminded me that the experience happens at x.convai rather than convai. Once I set browser preferences for x.convai to allow pop-up windows and microphone access, the problem was solved. I can now follow the published link and speak to my character and all the other sample characters. Hopefully this will solve the problem for my collaborators as well. I’ll check with them and let you know.
Unfortunately, the x.convai site settings that solved the problem for me didn’t solve the problem for my collaborators. They still cannot get any response from my character.
Hi - I’m working with toasterguy to try to resolve this issue. We have tried accessing the character via the Web SDK and I have confirmed the API Key and Character ID is correct using Curls to access the API. When I try to use an aiuthored, audio only portal to talk to the character it just times out acting like the page code location is not whitelisted. I have confirmed in the account that the page is definitely whitelisted and waited many hours in case there was a propigatation issue with the whitelisting. When I look at the network activity in the console, I see nothing and it just times out as if it was not whitelisted.
Hi - I just wanted to ping the thread to see what we should try to resolve this or if there are other logs that would be helpful to post for debugging. It definitely looks like some type of access permission issue and the site is definitly whitelsited and we have plenty of credits.
As mentioned last week, none of my company partners can interact with my published characters. They appear on the pixel-stream screen, but do not respond to voice or to text at all.
dave.g, who joined the thread, is one of my company partners, and has encountered this problem every time he tries to talk with my published characters. He provided details about his tests 4 days ago, and is checking to see if there is a response to those details.
I understand, but what your colleague described seems to be a different issue. If we focus on the problem shown in the video, as mentioned in my previous reply, please share the browser console log at the time the issue occurs so we can investigate further.
The account you’re logging in with doesn’t have access to that voice. We see this in the logs: “doesn’t have access to the voice 11LABS-AUTO-Neuro-Test-Voice-”. Could you please try with GCP or Azure voices instead?
We are using ElevenLabs for our voice and understand that is what is causing the access limitation. When we replace the voice with a stock voice, the application now works. I am now exploring using pixel streaming embeding to see if that will works for us for our POC (as mentioned as the other issue reported previously on the thread).
