Foundation model for isolating any sound using text, visual, or temporal prompts • Meta AI Research • Fully open source