Segment any sound with text, visual, or time prompts
SAM Audio is a unified model that separates any sound from any source. Use text ("dog barking"), visual clicks on video, or time spans to isolate specific audio. It unifies speech, music, and sound effect separation into one promptable model.