Suno has unveiled the sixth version of its music generator, developed in collaboration with major record labels including Warner Music Group, BMG, and Believe.
This new model is trained exclusively on licensed data. Journalists from TechCrunch have linked this decision to ongoing lawsuits from copyright holders against Suno. Just a day prior to the release of version six, the company acknowledged that it had been collecting data from YouTube videos.
The update features three distinct versions of the neural network:
- v6 — the main algorithm available for Pro and Premier subscribers, which adheres closely to prompts and produces consistent results;
- v6-wild — a model designed for creative experimentation (available through a paid subscription), which generates unconventional and surprising compositions;
- v6-mini — a fast, basic version accessible to all users for free.
In this sixth generation, developers have focused on providing precise control over compositions. New editing tools have been introduced:
- point edits: users can request changes to just the chorus or correct a single word in the lyrics, leaving the rest of the song intact;
- sample isolation: the neural network can isolate individual instruments, allowing users to cut a guitar riff from the middle of a track and overlay it with a different beat;
- mashups: the algorithm can combine vocals from one generated song with drums from another.
- multimodality: users can now create music based on uploaded images, videos, and voice recordings.
Suno plans to fully transition to version v6 soon, removing the older models from its platform.
Looking ahead, the company aims to launch new products to support real musicians. Artists will have the option to voluntarily provide their data for generation and receive a percentage from the platform, while fans can legally create tracks in their style.
It is worth noting that since September 3, Suno has implemented strict limits on track downloads.