Hi, after recent updates, when I publish my google ai studio app the live chat is not working it says: “Audio connection lost.” I have tried to fix it through the prompts but still it is not working. On preview, live chat works. but on the browser it doesn’t. I have reset the browser and I have used cognito window and I have logged out and logged back in, but still not working. This is so frustrating.
2 Likes
The “Audio connection lost” error on published apps (but not in preview) is almost always a microphone permissions issue in the browser.
Here’s the most common fix:
1. Check browser microphone permissions
When you access a published app via its public URL, the browser treats it as a different origin than the AI Studio preview. Go to your browser’s address bar → click the lock/info icon → check if microphone access is allowed.
2. Allow mic permissions when prompted
The first time you open the published app, the browser should ask for mic access. If you accidentally denied it, you need to reset it: go to chrome://settings/content/microphone (on Chrome) and remove the site from the blocked list, then reload.
3. HTTPS required
Make sure the published app is being served over https:// — browsers block microphone access on non-secure connections.
4. Check if it’s a recent regression
If this worked before and broke after an update, it might be a known bug. Check the “Known Issues with Stitch” pinned thread to see if it’s been reported.
Try step 1 first — in most cases that’s the issue.
1 Like
Thank you very much Ahmad for the explanations. The live chat was working fine and after the update, it doesn’t work on my Chrome browser on laptop or on my iPhone. the first 3 points that you mentioned are okay. Regarding the fourth one, I have not been able to find anything about the issue.
I have reproduced this multiple times now. Steps:
- Build a simple conversational voice agent app.
- Add a Gemini key in the Secrets panel.
- Test in AI Studio (works).
- Publish and test (doesn’t work - connection is never established or drops instantly).
Debugging with the AI itself suggests two issues:
- A conflict between the Gemini SDK’s default WebSocket URL and the AI Studio proxy interceptor, which resulted in a double slash (//ws/) in the connection URL.
- AI Studio does server-side string substitution on the fly, while Cloud Run just serves the raw compiled JS bundle containing the literal string “MY_GEMINI_API_KEY”, where “MY_GEMINI_API_KEY” is the name of the secret holding your gemini key.
It is relatively easy to get a workaround for issue 1 but all efforts to get the agent in AI studio to fix issue 2 have failed, consuming a lot of tokens in the process. It alternates between fixing the published app (and breaking it in Studio) and vice versa.
Some help or guidance on this from the AI Studio team would be appreciated. I have had to cancel some demos because of the issue. I had never previously published a voice agent so I don’t know if the issue is due to the latest updates, but the net effect of this issue is that a voice agent cannot be published successfully. If anyone has been able to do so recently, please advise.
1 Like
Hi Louis, thank you for the explanations. The live chat was working fine and is still working on some apps that are not yet migrated. But for this new update that I am developing, after the migration, the live chat start working. I am developing a language learning app and live chat is the main feature of the website.
Hi Barry, thanks, are you saying here that your app was not working but now is, after the migration - or that it is still broken?
Same issue tried many things planning to switch to different api model now
I am experiencing this exact same issue. Has anyone figured this out. Would of thought Google would have fixed, since this original post was from 24 days ago…