We tested various AI music generators

We often discuss image and video generators, but let’s shift our focus to music generators — a burgeoning field in the AI landscape. With a plethora of options available, we decided it was time to put some of these tools to the test to gauge their evolution and what they currently offer.

These generators operate by analyzing extensive datasets of music, learning from a variety of styles and compositions. Users can specify parameters such as genre, tempo and mood, guiding the AI to produce music that aligns with these preferences.

As filmmakers and musicians ourselves —fact, check out our Spotify—, we find these tools interesting, as they help quickly sketch out song ideas or create simple tracks for social media videos or others.

COPYRIGHTS AND LICENSING WITH AI-GENERATED MUSIC

Now, before we go on, keep in mind that the rise of AI-generated music obviously brings up significant questions regarding copyrights. So, it’s essential for users to understand the terms and conditions of usage and licensing, as they can vary greatly between different services. Some platforms might retain rights to the music created, while others may offer complete ownership to the user. Some allow you to monetize the song, but only if you have paid plans. Always review the licensing agreement carefully to ensure that you retain the rights needed for your intended use, may it be ads, content creation or even short films and movies.

TESTING AI MUSIC GENERATORS WITH THE SAME PROMPT

To compare the capabilities and outputs of different AI music generators that you can try out right now, we used the same prompt across different platforms. This experiment highlighted the strengths and creative diversity of each tool.

Prompt: “Make a grunge song, very passionate and depressed, the likes of Something in the Way. Make it about self-doubt, control and aliens”

UDIO - As the newest kid on the block, UDIO is breaking the internet. Despite being in its beta testing phase, the results it produces are quite impressive. It effectively captures the intended genre and the voices are convincing. It generates two songs per prompt, each approximately 30 seconds long, complete with lyrics. The songs can be extended, if needed.

Currently, the control over the output is limited; you can only adjust a few parameters such as prompts, tags, whether the lyrics are custom or generated, and if the composition should be instrumental.

Vast Unknown

0:00

Lyrics generated

SUNO – We first tested it last year and it seems to have had a nice evolution since then. Today, it generates two songs per prompt, each with two verses and a chorus. Additionally, it creates lyrics—which you can customize or you can put your own—and an accompanying image for your song. It’s user-friendly, though it lacks extensive parameters to control the final output, similar to UDIO. The voices are decent and it generates some intriguing ideas.

Shadows of the Mind (1)

0:00

Lyrics generated

SONAUTO – This tool is quite straightforward; you can use a prompt or even a song as a reference, and that's it. It generates three songs per prompt, complete with lyrics. However, the quality leaves much to be desired—it's as if we're dealing with a band that needs more rehearsal. The "singers" require significant improvement and the overall sound is pretty messy. Despite these shortcomings, it does manage to create a song within the specified genre, and it's also fun to hear AI fail. :)

sonauto-generation

0:00

Lyrics generated

BEATOVEN – This tool is primarily designed to create moods and atmospheres rather than full songs. We conducted some tests with it a year ago and noticed little evolution in its capabilities since. While it offers several parameters to control aspects like emotions, instruments, tempo and genre, it hasn’t fully satisfied our requirements or lived up to our expectations.

Disclaimer: they don't allow downloads unless you pay a subscription, so here's the one we created a year ago - as said, the concept is similar, more atmospheric than songs per se.

SOUNDRAW – This one is very different from the others. It doesn't require a text prompt; users simply set the length, tempo and genre. It generates—or rather, spurts out—a multitude of song ideas that sound more like MIDI tracks, serving as a foundational base for further creative development.

The tool also allows users to “shorten intros, rearrange choruses, and personalize your song structure”, as described on its website. Although it doesn't support adding vocals for this genre, our tests with trap music revealed it occasionally inserts brief 'hey' sounds—not full singing voices. We recommend keeping an eye on this tool if you're a musician. However, for filmmakers, it might not be the ideal choice.

Disclaimer: they don't allow downloads unless you pay a subscription.

Web interface - how you can modify your music.

SPECIAL MENTION: AIVA of Nvidia - it offers the ability to create specific and customized music. Unlike others, it doesn't accept text prompts. Instead, you can create a song based on a style, a chord progression, step-by-step adjustments, or musical influences.

It's more complex than the others, and we plan to explore it further. However, for the purposes of this article, it's not included in our main comparison. What we've observed so far is its limitation in recognizing the 'grunge' genre. It's also worth noting that this tool seems to be particularly suited for producers and musicians, rather than filmmakers or general content creators.

AIVA web interface

We also explored other tools like Mubert and Stable Audio, though these didn't quite capture our interest or provide the fun results we hoped for. And there's still a wealth of AI technology out there to explore, such as Soundful and the upcoming MusicLM from Google.

In conclusion, much like other types of AI generators, those that generate music can be both fun and useful. It’s essential that we harness them wisely to ensure that creativity flourishes without stagnation, while also respecting the rights of all creators involved. How we use these AI tools will significantly shape the future of art. What is certain is that these music generators will enable more people to explore their musical potential, paving the way for a new breed of artists to emerge.