Suno releases v6 music model trained with record industry partnerships

Suno has released its latest music generation software, designated version 6. This update marks a significant shift for the company as its first model developed with support from the music industry.

Data and Model Architecture

The training data for version 6 was built from the ground up using a new dataset. This includes content licensed directly from partners Warner Music Group, BMG, and Believe, alongside user data.

The new release splits into three distinct model variants:

• v6: the standard model offering improved genre adherence.
• v6-wild: designed to introduce unpredictable elements and happy accidents.
• v6-mini: a lightweight model available for free use that focuses on faster creation.

Capabilities and Limitations

All three versions show a marked improvement in understanding complex musical genres, accurately capturing superficial hallmarks that previous iterations missed. However, the system struggles to produce deliberate acoustic flaws. Attempts to generate off-key vocals, out-of-tune pianos, or rhythmically haphazard performances consistently yield harmonic and rhythmic perfection.

Additionally, the software continues to exhibit artificial artifacts, particularly in vocal tracks, which can sound harsher than those produced by version 5.

Workflow Updates