-
-
Notifications
You must be signed in to change notification settings - Fork 191
Open
Description
Summary
Currently HAPI's voice assistant only supports ElevenLabs Conversational AI as the voice provider. It would be great to add MiniMax Speech as an alternative option.
Why MiniMax?
- MiniMax Speech-02 has topped global voice evaluation benchmarks, outperforming both OpenAI and ElevenLabs (source)
- Chinese voice quality is significantly better — character error rate of 2.25% vs ElevenLabs' 16%
- Cost: ~1/4 of ElevenLabs pricing
- Low latency: Speech 2.6 achieves <250ms end-to-end, comparable to ElevenLabs
- Features: 300+ system voices, voice cloning from 5-second samples, 40+ languages, emotional control
- MiniMax is already widely adopted by platforms like LiveKit, Vapi, and Pipecat
Proposed Change
Add MiniMax as a configurable voice provider alongside ElevenLabs, e.g.:
export MINIMAX_API_KEY="your-api-key"
export HAPI_VOICE_PROVIDER="minimax" # default: "elevenlabs"
hapi hub --relayContext
For users in China or those working primarily in Chinese, MiniMax provides a much better experience and is more cost-effective. Having provider choice would make HAPI more accessible globally.
References
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels