Details
Users input a text description (e.g., genre, mood, artist reference, or custom lyrics), and Udio's generative AI models produce two complete tracks of approximately 30 seconds each, which can be extended in 30-second increments up to 15 minutes. The AI handles lyrics, melody, vocal synthesis, arrangement, and mixing. The specific model architecture has not been publicly disclosed, though third-party analysis suggests it likely combines a large language model for lyrics with a diffusion or transformer-based audio generation model. The platform launched publicly in beta on April 10, 2024 and is accessible globally via web browser and a mobile app.
Have evidence about Udio's AI practices? Submit a report.
Report a Sighting →