Content
summary Summary

Update from March 23rd, 2024:

Suno v3 is now generally available, and you can try it for free. According to Suno, the new model can produce "radio-quality music" up to two minutes long. Suno says the audio quality has been improved over the alpha, and there are more styles and genres to choose from. The model follows text prompts more closely, with fewer hallucinations and better song endings.

To prevent misuse and another "Fake Drake", Suno says it has "developed proprietary, inaudible watermarking technology that can detect whether a song was created using Suno."

Original article dated February 24, 2024:

Ad
Ad

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.

Recommendation

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.

Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.

Suno offers its song generation service through 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.

Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • 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.
  • The new version offers better audio quality, more expressiveness and speed, supports more languages and instrumentals, and allows songs to continue at any point.
  • During the trial period, users get 300 free credits to experiment and can switch between V2 and V3 Alpha. Suno AI plans to improve the model based on user feedback.
Sources
Online journalist Matthias is the co-founder and publisher of THE DECODER. He believes that artificial intelligence will fundamentally change the relationship between humans and computers.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.