Suno Launches V6, AI Music Model with Licensed Content

Key Takeaways

1Suno begins rolling out v6, its AI music model trained with licensed content from partners like Warner Music Group, BMG, and Believe

2v6 comes in three variants, including a free version, and introduces chat-based editing as well as multimodal prompts

3The exact composition of the data remains uncertain, and reported trials highlight limitations on generating "natural imperfections"

Why it matters : v6 marks an evolution in the collaboration between AI music and the recording industry, while raising questions about data transparency and the creative capabilities of AI.

Suno begins rolling out v6, its new generation of AI music, and plans to retire its older models. Trained with licensed content from partners like Warner Music Group, BMG, and Believe, v6 comes in three variants and introduces chat-based editing as well as multimodal prompts. The company states that it has used a new dataset, but the exact composition remains partially uncertain. Reported tests also highlight limitations in generating "natural imperfections."

Gradual Withdrawal and Uncertainties About Data and Output

Suno plans to phase out its previous models as v6 is deployed, a rollout that starts today. The exact composition of v6's training data is not fully clarified, particularly regarding the absence of content obtained through dubious means. In limited tests, v6 retains certain unnatural imperfections typical of AI-generated music, with voices exhibiting more hard-edged artifacts than the previous version, v5. The same tests report that v6 fails to produce "natural imperfections" such as slight dissonances or out-of-tune instruments, and that queries like "monotone voices" or "no drums" are not accounted for, even for a deliberately out-of-tune piano.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Claimed Training on Licensed and "User" Data

Jack Brody from Suno indicates that v6 has been trained from the ground up with a new dataset, different from that used for previous models. He specifies that this training includes licensed content from partners like Warner Music Group, BMG, and Believe, as well as user data. v6 is presented as Suno's first model to benefit from support from the recording industry.

Three Variants, from Free Quick to "Wild" Mode

The v6 lineup includes three models: v6, v6-wild, and v6-mini. The mini version is available for free and aims for quick generation with low resource consumption, but its results are simpler and often exhibit more characteristic AI artifacts. v6-wild is described by Jack Brody as more unpredictable, designed to encourage "happy accidents" and "natural imperfections." All three models are reported to be significantly better at understanding musical genres, with tests showing that v6 has successfully reproduced elements specific to styles like hyperpop or krautrock.

Chat-Based Editing and Multimodal Prompts

v6 now allows users to edit specific parts of a song using simple language in a chat, without needing to regenerate the entire track, whether it’s a guitar line or a word. The model can also merge multiple elements from the user's Suno library to create new compositions. Prompts are no longer limited to text: it is now possible to generate pieces from images, videos, or other audio files.