Despite concerns over potential copyright infringement, Forsgren insists that Riffusion cannot replicate famous artists or songs. The team is currently focused on expanding and developing new generative AI products, with potential plans to collaborate with artists like The Chainsmokers. The upgraded Riffusion is powered by an audio model trained from scratch by the six-person team, using spectrograms of music tagged with relevant terms.
Key takeaways:
- Developers Seth Forsgren and Hayk Martiros have commercialized their AI-generated music tech project, Riffusion, which has closed a $4 million seed round led by Greycroft.
- The new version of Riffusion allows users to describe lyrics and a musical style to generate “riffs” that can be shared publicly or with friends.
- The Riffusion team trained an audio model from scratch, which uses spectrograms or visual representations of audio to generate music based on user prompts.
- Despite potential copyright concerns with AI-generated music, Forsgren asserts that Riffusion was not trained to recognize famous artist names or songs and cannot replicate them.