Google state-of-the-art conditional music generation model transforming descriptive text into 24kHz audio
Free experimental access via Google AI Test Kitchen to generate and download audio clips
MusicLM is Google Research groundbreaking foundation model for conditional music synthesis. Rather than parsing mere genre keywords, MusicLM understands nuanced, expressive descriptive sentences (e.g., "slow-tempo techno track with a deep hypnotic bassline, warm 80s synth pads, and dawn birdsong") and generates coherent 24kHz audio tracks lasting minutes. Trained on an extensive corpus of diverse musical recordings, it captures intricate instrumental timbres, melodic structures, and dynamic transitions via Google AI Test Kitchen.