Audiogen

Audiogen is an applied research lab and web app for AI music creation. Its small team of researchers, engineers and designers spent about two and a half years building a generative music model and a new interface for working with it, because they were unsatisfied with how existing music models were deployed and used. The model stack has several named parts: Aphrodite, an audio codec running at roughly 3 kbps, Apollo, a diffusion transformer, and Virgil, a synthetic audio captioner. On top sits a browser based Generative Audio Workstation, a rethought version of the timeline in a digital audio workstation. Inpainting, extending, remixing and multi track or stem editing are designed to feel like painting on the screen, with the goal that the creator still finishes a song with a sense of authorship. The site includes demos, a film, a community area, and sign up and log in pages. It is aimed at hobbyists and musicians who want to make music with AI in a beginner friendly environment, and at people curious about how generative music models are built. The lab describes the effort as a shoestring budget project with a small team. The site publishes no pricing, and no pricing page could be found, so costs or usage limits are unknown beyond creating an account. It is a different product from Meta's AudioGen model, which is part of the AudioCraft research code base.