Question

You are building a lightweight app in Azure AI Foundry that lets warehouse workers ask questions hands-free. Workers speak a question into a headset, and the app should reply with a spoken answer using a deployed multimodal model. When you configure the request to the multimodal model, which setting is essential so that the model returns an audio response the app can play back?

Accepted Answer

Specify the audio output modality (and a voice) in the model request so the response is returned as generated speech

Answer

Set the temperature parameter to 0 so the spoken answer is deterministic

Answer

Enable custom speech training to improve recognition of warehouse terminology

Answer

Increase the max_tokens value so the response is long enough to be spoken