diff --git a/fern/pages/speech-to-text/universal-streaming/select-the-speech-model.mdx b/fern/pages/speech-to-text/universal-streaming/select-the-speech-model.mdx index 0588e1d0..9c54b9f8 100644 --- a/fern/pages/speech-to-text/universal-streaming/select-the-speech-model.mdx +++ b/fern/pages/speech-to-text/universal-streaming/select-the-speech-model.mdx @@ -38,6 +38,10 @@ For detailed setup and configuration of Universal-3 Pro streaming, see the [Univ For detailed setup and configuration of Whisper streaming, see [this page](/docs/streaming/whisper-streaming). + +**Building a voice agent?** We strongly recommend **Universal-3 Pro Streaming** (`u3-rt-pro`). It provides the best accuracy for short utterances, entity recognition, and turn detection -- all critical for conversational AI. Combined with proper prompting and turn silence configuration, U3 Pro delivers significantly better out-of-box results for voice agent use cases. See our [Voice Agent Best Practices](/docs/voice-agent-best-practices) guide for complete setup instructions. + + ## End-to-end example You can select a model by setting the `speech_model` connection parameter when connecting to the streaming API: