We use minimax for custom voice via api. We get this error often. Between 20-50% of the calls are failing. Does Vapi have a plan to fix this because we need the low latency TTS. 11labs is not fast enough. We are losing many valuable calls due to this problem. Please guys.

Thank you for sharing the details. To help diagnose the frequent "minimax voice failed" errors, could you please specify if you see any particular error codes or messages in your Vapi API or call logs when these failures occur? This will help us pinpoint whether the issue is related to authentication, timeouts, or audio format in your custom TTS integration. For reference, you can review and debug TTS errors using the [Vapi dashboard logs](https://docs.vapi.ai/debugging) and ensure your custom TTS endpoint meets all [integration requirements](https://docs.vapi.ai/customization/custom-voices/custom-tts). Sources: - [Debugging voice agents](https://docs.vapi.ai/debugging) - [Custom TTS integration](https://docs.vapi.ai/customization/custom-voices/custom-tts)

I will only have access to the logs on Monday, but here are some examples of the most recent error calls: 019d0590-99a3-7ee5-bfc0-b2fac0317bfb 019d058b-e842-7ddb-a5ff-cc9c8ad6705c

Hi, The quickest way to reduce your call drop rate is to configure fallback TTS voices on your assistant. Currently, when Minimax fails, the call may drop but with a fallback in place, the system will automatically switch to a backup voice instead. You can set this up via the API by adding a

fallbackPlan

with alternative voices in your assistant’s voice configuration. Even adding a single fallback option (for example, a Deepgram or PlayHT voice) can significantly improve reliability and prevent interruptions during intermittent issues.

Hi Oshi. Are you saying that with fallback TTS the call will not drop and vapi will switch to fallback tts mid-call? There are not any TTS that are even close as fast as Vapi. We are in SA so we really need a super fast latency for lowest total round trip latency. We are 250ms away from you basically.

Hi, to provide you more brief, your original call failure was due to MiniMax-side issue. The quickest way to reduce your call drop rate is to configure fallback TTS voices on your assistant. When Minimax fails, the system will automatically switch to a backup voice instead of dropping the call.

Thanks, but we can only use Minimax due to the latency. Do you know which of your voice providers are second fastest after minimax?

Hi there, Thank you for your patience while we investigated this. The reason your 20-50% of the calls are failing because of 429 error. This indicates that the API rate limit has been exceeded. Since this is coming from the MinMax provider, we recommend checking with them regarding your rate limits and usage. Also, on the screenshot you shared, for custom LLM setup VAPI don't provide fallback plan configuration because of this, the system is showing an error.