DiffRhythm AI

DiffRhythm AI generates complete songs with vocals and accompaniment in seconds. Enter your lyrics and describe the style you want to get a finished track up to 4 minutes long.

Text-to-song generation

Provide lyrics plus a genre, mood, or reference, and the model synthesizes both the vocal line and instrumental backing. The output is a cohesive composition rather than separate fragments.

Latent diffusion music model

DiffRhythm AI is based on a research model developed with participation from ASLP Lab. Its architecture is designed for end-to-end generation—from a text description to final audio—without manual track assembly.

Who it’s for

DiffRhythm AI can be useful for:

• Songwriters who want to quickly test ideas and drafts
• Producers exploring AI-assisted music creation
• Social media creators who need original music with vocals
• Anyone experimenting with AI music workflows