The company’s latest release, v6, is the first AI music model to come from Suno with assistance from the record industry. According to the company, it drew on licensed material from Warner Music Group, BMG, and Believe, along with user data. Jack Brody, speaking for Suno, said that v6 was “trained from the ground up, with a new set of data that does not include the same data that our previous models were trained on.”
There is no clear confirmation yet about whether v6’s training material contains any dubiously obtained content.
Three models, one release
The Suno v6 release arrives as a trio of distinct models, with v6, v6-wild, and v6-mini each taking their own path. The v6-mini comes at no cost to anyone who wants it, built for quick, light creation that demands little from its host system. Its output stays simpler than the other two, and markers of its artificial origin tend to show through more often.
The v6-wild model is meant to be more unpredictable. Brody said it is designed for “happy accidents and natural imperfections.” In testing, The Verge found it hard to notice a difference between wild and standard v6.
All three models are dramatically better at understanding genre. Prompts for “hyperpop” or “krautrock” that missed badly on older models now nail the superficial hallmarks of any genre thrown at them.
What v6 still can’t do
The notes beyond the “natural imperfections” still cannot be reached. Just as with v5, efforts to produce off-key, out-of-tune, or slightly dissonant sounds consistently came up short.
Suno’s v6 holds fast to harmony and rhythm no matter how much force is brought to bear upon it. A request for “monotone vocals” and “no drums,” was ignored outright. Demands for “out-of-tune piano played haphazardly, with no regard for key or rhythm” produced nothing resembling the desired result.
“happy accidents and natural imperfections”
V6 keeps the artificial rough edges of AI music intact. The singing parts in particular appear to generate more rough-edged artifacts than v5 did.
New ways to edit and create
The Suno v6 update changes how users work with the software, adding behind-the-scenes improvements that alter the experience of interacting with it. Now, you can make edits to parts of a song directly inside a conversation window, using natural language. That means changing a single guitar line or a single lyric no longer forces a full regeneration of the whole track.
New creations can be built by combining several pieces from your Suno library, with the model joining outputs into something fresh. The old limit of text descriptions alone has now passed.
- v6 can create tracks based on images
- v6 can create tracks based on video
- v6 can create tracks based on other audio
The rollout of Suno v6 has begun, and the company has said that its older models will eventually be retired.
Source: theverge.com
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.