Suno is one of the leading artificial intelligence-based music generators, allowing users to create entire songs with vocals using only text prompts.

In the year since its initial launch, it has rapidly improved, extending the duration of songs, adding new community features, and even allowing users to create singers.

Version 4 will take things to a new level. The next generation model will be released “soon.” Still, I have seen several examples of its output on social media from Suno employees and those granted early access, and it is impressive.

Suno has an iPhone app that allows users to create songs from images, which will be included when v4 is released; if you want to get started with Suno, see How to Use the AI Music Generator.

I have not personally tried Suno v4, purely from watching videos shared on social media, but the vocal quality and overall production tone seems to be leaps and bounds better than anything I have seen.

Other new features may be added to v4, but for now, purely on sound alone, it will be an upgrade that everyone will want to try. Many of the sample tracks I heard made me look twice to see if they were really AI.

One area Suno has always excelled in is country, but the vocals always sounded like someone had applied too much autotune. This imolivercom sample sounds like it belongs on a top Spotify playlist.

The combination of music and video would be huge, especially now that we can lip-sync. Below is an example from X by AI creator DAViD that shows just how good the audio part of Suno v4 can be.

It begins with a monkey and a monkey talking about Suno in general, then sings along to music created using v4. He animates with Viggle and says, “The progress in song generation and vocals with V4 is miles ahead of V3. This is just the beginning!”

.

So far, most of the sample clips are only a few seconds long, but more than enough to see a marked improvement in vocal tone.

The Suno team has released a v4 sample of a soulful song with a John Legend vibe; Suno has put a lot of effort into making sure users can't duplicate the sound of the real artist, and the sample is a great example of this.

This sample is what first got me excited about the new model, as it shows a level of vocal fidelity that even human singers struggle with.

Finally, back to imOliver, who answered a question from a follower about handling heavy songs. This is a new metal track called Stone that I made in v4.

I don't think it's as good as some of the other examples, but I'm not much of a metal, nu-metal, or heavy music fan, so maybe that's my personal bias. I think v3.5, which excelled at heavier music, came close to this. [AI-generated music has progressed rapidly, going from not being able to generate vocals at all to being able to create custom singers across multiple tracks that sound like humans in a year.

Not everyone will be a fan of AI-generated music. There is an ongoing court case over the use of commercial music in training data, and musicians are concerned about the impact it will have on their income. [33] [34] I think both worlds can exist, and in fact, we could be on the verge of a new golden age of music, with new genres, mash-ups, and styles emerging faster than ever before due to human musicians using AI, and experimentation becoming cheaper all at once.