Stability AI, the creator of Stable Diffusion, has announced a new family of audio models called Stable Audio 3.0. The company claims that the most advanced model in this series is capable of generating professional-grade music lasting over six minutes. The new family includes four models: small SFX, small, medium, and a large model with 2.7 billion parameters. This is reported by Techcrunch.com reports .
The small models are designed to generate up to two minutes of sound and music directly on devices. The medium and large versions can create full compositions of 6 minutes and 20 seconds. This is twice the capacity of Stable Audio 2.0 released in 2024. Stability AI has released its small and medium models with open weights, allowing users to modify and use them for personal purposes.
The largest model is only available via API and paid services. Additionally, companies with revenue exceeding $1 million are required to obtain a corporate license. Currently, giants like Google and ElevenLabs are also developing their own music generation tools. However, as legal proceedings involving Suno and Udio have shown, copyright and licensing issues are critical to the future of these services.
Stability AI stated that its new models were trained on fully licensed data. Last year, the company signed partnership agreements with Warner Music Group and Universal Music Group. The startup is also developing a new suite of products for professional musicians, and former Universal Audio executive Ethan Kaplan has joined the team to lead this direction.
Read “Zamin” on Telegram!
