MLModel, not as a ROS intelligence layer. Audio goes in, a JSON transcript comes out, and the same model can be called from POST /api/v1/mlmodels/{uuid}/run or a workflow CALL_MODEL node.
The default catalog model is Whisper Large v3:
Models
Speech to Text
Transcribe robot audio with Whisper MLModels in workflows, REST calls, or edge workers
stub
Cyberwave exposes speech-to-text as an
The Whisper Cloud Node returns:
Use this when a microphone driver uploads audio and a workflow needs to pass the transcript into a downstream controller policy, planner, or human-in-the-loop step.