Zero-shot

Zero-shot refers to an AI model performing a task, or handling a language, speaker, or condition, that it was not explicitly trained on, generalizing from what it learned rather than needing task-specific examples. Zero-shot capability is valuable in audio and speech because a model can work on new voices, languages, or content without retraining. It contrasts with approaches that require labeled examples for every new case.