Plugin 3.6.8-beta3, Claude 4.0 Opus and GCP voice giving weird error

Working with Core AI: Claude 4 Opus, just received this error. On Plugin 3.6.8-beta3

API Error: Error code: 429 - {‘error’: {‘message’: ‘Provider returned error’, ‘code’: 429, ‘metadata’: {‘raw’: 'meta-llama/llama-4-maverick is temporarily rate-limited upstream. Please retry shortly, or add your own key to accumulate your rate limits: OpenRouter ', ‘provider_name’: ‘Groq’}}, ‘user_id’: ‘user_31TqtEYL4wcZq2Kdyo7Ha518dSz’}

image

Character ID 60e087de-76af-11ef-80ce-42010a7be011
Session ID c0c01a2ee0c3f3f266e33f3a2d76bc11

Never had this error before!

1 Like

This is how the Digital Human is speaking with Convai Plugin Version 3.6.8-beta 3, no matter which Core AI I select, it is giving me gibberish information of API key error or speaking as above image. This is new.

Also, I noticed that the Error 14 specifically presents itself when using Claude 4.1 Opus as the Core AI Model.

For now, could you please use a different LLM? Gemini or OpenAI

Claude 4 singer serves to be working however, with GOT Ace Gemini, I’m getting the digital human speaking API Error message. This is new suits never saw these before.

1 Like

Do you mean the speak tags shown in the message? I’ve informed my teammates. They will investigate and get back to you.

Hello Jeasy,

Are you still having problems?

1 Like

I just changed the Core AI model to Claude-4-5-sonnet (beta) for now to test, and it seems to be working… The Error 14 showsup after when using Core AI with the Claude 4.1 Opus.

However, claude-4-5-sonnet (beta) doesn’t have as good emotions as Claude 4.1. It works in the playground but the Unreal Engine Packaged Build shows the Error 14 constantly. Also it take long for responses. Too much lag.

Just tried switching to Claude 4.1 Opus to test with the build, and Error 14 is back. Switching back to Claude-4-5-sonnet (beta) for now.

Vivience20_20251118_104001_Default.log (69.1 KB)

Just to clarify again: Error 14 is a separate issue, and this thread is focused on a different error. Please try our beta package for Error 14.

I will start testing the new beta plugin next week. and report back in a new thread. I believe the errors and speak tags, as well as timeout issues, are all interconnected with the Core AI model, particularly for the current plugin version 3.6.8. Thanks, Kaan.