Skip to main content
The default speech recognition works for most conversations. Add asr_config to Create Session only when you want to use nova-3 for an English conversation.

1. Switch the provider

Merge this into your working Create Session request; keep its image, voice, and output settings. For multilingual conversations, leave asr_config out and use the default doubao.

2. Listen to the result

Create a new session and try a few English phrases from your application. Check how it handles names and sentence endings. The provider changes how user speech is transcribed; the avatar’s reply language still comes from its speaking instructions.