Suno has launched a new generation of its music models called v6 and immediately announced that it will gradually shut down the older versions. Within a few weeks, the entire platform will switch exclusively to the new models. The company developed them together with the music labels Warner Music Group, BMG, and Believe, and emphasizes that v6 was not trained on the same data as the previous versions.
Three models instead of one
For the first time in Suno’s history, users are not getting a single model, but three at once. Each model is aimed at a slightly different type of user.
The standard v6 is the flagship model for Pro and Premier subscribers. It is intended to be reliable and precise, which in practice means that if you know what you want, that is what you will get. The v6-wild version is the exact opposite. It is also available only to paying users and is deliberately unpredictable, producing wilder, rougher, and less polished results. The company offers it as a source of ideas that can be developed further—for example, by taking a finished idea back into the standard v6 model and refining it there. The third version, v6-mini, is a faster, lightweight option for everyone, including people without a subscription. Suno claims that it is better and faster than any free model on another music platform.
According to the company, the main improvement across all three models is their understanding of how musicians talk. Vocals, instruments, song structure, mood, and references to other styles—the new model is supposed to understand all of these better than before.
What you can now do with music
The biggest advance concerns editing finished tracks. Until now, if you did not like the chorus, you had to accept that regenerating it would also change the rest of the song. The new Suno can change just one section and leave the rest untouched. All you have to do is write a sentence asking it to rewrite the chorus so that it is sung by a choir. Lyrics can be edited in the same way. If a single word in one verse does not sound right, you can replace it without having to remake the entire track.
There is also a new option to create a mashup from several sources at once and describe in a single sentence how they should work together. Suno gives the example of taking vocals from one track and drums from another, adding new lyrics, and requesting the whole thing in an ’80s synthwave style. Similarly, in a single step it can take a sample from a specific second of a recording, extract something such as the guitar, and build a new rhythm around it.
The model can also handle prompts that describe a feeling rather than a genre. According to the company, a sentence asking it to make a song that sounds like midnight on a rooftop is enough to provide a starting point. You can begin with virtually anything, specifically text, a voice message, a photograph, or a video.
This time, only licensed music
This is the reason v6 is attracting attention. Suno developed the models with record labels as partners and says it did not use the data on which it built its older versions for training. The company reached an out-of-court settlement with Warner last November, concluded an agreement with BMG last month, and announced its partnership with Believe the day before unveiling v6.
Before launching the new models, the company also introduced checks of uploaded files and lyrics designed to detect material used without authorization. It also promises greater transparency about where and how artificial intelligence has been used in music.
But the lawsuits have not ended
Changing Suno’s training data has not erased the company’s past. It is still fighting lawsuits brought by the labels Sony and Universal Music Group, and singer-songwriter Jason Isbell has also sued it. Its own users filed a class-action lawsuit over a data breach, accusing the company of prioritizing profit over security.
The timing is all the more striking because, one day before announcing v6, the company admitted that it had obtained audio from videos on the YouTube platform to train its models. At the same time, it questioned whether Universal and Sony even have standing to sue over the matter.
In any case, Suno is trying to strike a conciliatory tone. Last month, it announced that it would add watermarks to generated tracks and introduced new download limits based on account type. Chief Product Officer Jack Brody said that the platform can help address streaming fraud and the bulk uploading of low-quality content to distributors, but in his view, monitoring what appears where is primarily the responsibility of distributors and streaming services.
What comes next
The next phase of the plan revolves around specific artists. Suno is preparing a program that artists will be able to join voluntarily, allowing their tracks to be used for AI-powered features in exchange for payment. Fans could then remix music by their favorite artists directly in the app. Brody explains that labels and partners are looking for new revenue streams for their artists, and this is intended to be one of them.
Suno has had a very busy summer. In mid-August, it released Studio 2.0, a major overhaul of its browser-based music editor that can work with MIDI, audio effects, and a built-in synthesizer. In early September, it added an improvement that allows the chat bar to understand tempo and handle speed changes.
Those who like breaking tracks down into their component parts received a new stem-separation feature in June, allowing them to choose exactly what they want to extract from nearly one hundred instruments. In July, lyric editing was redesigned, with the addition of full-screen mode, autosave, and the ability to edit verses using a plain sentence, such as asking for a particular line to be funnier. Voices, a feature that can transfer your own voice into tracks, arrived on mobile phones in August and can also be tried by people without a subscription.
Suno certainly has the money for these projects. According to PitchBook data, it has raised more than $819 million from investors in total.
Sources: suno.com, suno.com and techcrunch.com
