Suno’s new music AI model blurs the line between generated and human-created songs


Suno AI has unlocked access to V3 Alpha for its Pro and Premier subscribers and invites users to test and provide feedback on the new version of its music generation model.

US start-up Suno specializes in AI audio generation from text. The current music model v2 can already generate some impressively authentic songs. However, if the first demos are anything to go by, the new version v3 will go way beyond that.

Video: v3 prompted by Suno AI

Video: v3 prompted by Suno AI



V3 Alpha offers better audio quality, is more expressive, and generates faster. It also supports more languages and instrumentals.

In addition, songs can be resumed at any point instead of only at the end of the generation. The new model can also generate unusual sounds, such as these incantation chants.

“V3 Alpha creates more realistic, authentic music, and we believe it has the potential to redefine state-of-the-art for generative music,” writes Suno AI.

According to Suno AI, there are still issues with keys, BPM, and mixing and mastering songs depending on the prompt. The new model can generate songs up to two minutes long, but has hallucinations especially with shorter prompts. V3 Alpha takes more creative liberties with song structure and sometimes requires different prompts than V2 to achieve optimal results.

Users can switch between V2 and V3 Alpha during the test phase and receive 300 free credits to experiment. Suno AI wants to improve the model based on user feedback.


Discord and a website. The tool also generates lyrics for songs. All you have to do is enter a general theme.

Music styles and elements such as tempo can also be specified via text input. For copyright reasons, Suno AI does not support the input of artist or band names.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top