Suno vs. Udio: The Battle of Beats That’s Shaking the Music Industry!
In recent months, Suno has made a huge splash in the news as an AI that generates music. Their case was somewhat similar to when ChatGPT amazed tech world journalists and bloggers.
The news was abuzz with claims that composers were doomed, and everyone could hang up their guitars and synthesizers. Some music rights organizations have been communicating the musical apocalypse on Meta platforms ever since, highlighting Suno in paid advertisements.
Then something happened at the beginning of 2024; a new contender appeared in this segment, Udio. Suddenly, the tones changed, many declared a new king, calling it the Suno-killer, though most of these were articles optimized for clickbait consumption.
In this test, we seek answers to two questions:
- Have composers and music producers really become obsolete?
- Is there really a new king of generative music-making?
Disclaimer:
It’s pointless to argue about taste, although in the arts (even AI-generated art), personal taste always influences what we watch or listen to. As a result, while I strive to be as objective as possible, I’m sure my personal opinion/taste will somehow seep through.
That was enough foreplay, for a new relationship not for a blog post so, let’s dive into our main topic:
What Can They Technically Do?
With Suno, the v3 model was recently made available to all users (it was in beta for a short time before being implemented), but the sound quality was only increased to 192 kbps from the previous 128. Consequently, the spectrum is quite lacking, which already excludes many uses, although with some post-processing, there are plenty of opportunities (we’ll return to this later).
Udio sensed a gap in Suno’s armor and aimed right there, as they made the biggest splash in terms of sound quality right from their beta release. With Udio, you get songs in 320 kbps quality from the first moment.
However, the path from prompting to finished music is quite different for the two development teams.
While Suno tries to create a song in as complete a form as possible, Udio compiles music from 33-second segments. In terms of the end result, similar things can be achieved, but with Suno, you have to plan ahead for a song’s structure, while Udio gives you relative freedom to make changes on the fly.
Which is better, I think, depends on personal workflow, as some people think in complete songs before sitting down at the computer, while others use the “try it out and see what happens” method. Of course, we can expand and continue the primarily generated music in both cases.
If we build meticulously, in pieces, then Udio works better for us, while those who sit down with a complete concept might prefer Suno.
We must mention a few flaws for both platforms.
Suno:
With Suno, if you want to generate an intro for a given song afterward, you may encounter serious problems, so it’s worth planning ahead and always inserting the [Intro] prompt element at the beginning of the prompt.
Previously, both platforms had a remixing option, but this was “updated out” of Suno v3.
Udio:
With Udio, you can create an intro afterward, but since it operates in 33-second units, it will create the intro in this length, regardless of whether you need 5 seconds or just a drum fill.
With Udio, there’s also a problem if you give it a full song lyric at once, as the singer/rapper will rush it to fit within the 33-second limit.
Styles, Created Music:
Here, taste will definitely come into play, and most comments related to the article will likely be about this. As a fan of more popular music, I find that Suno serves my taste better in terms of instrumentation and structure, while country music, experimental songs, and typically „indie-style” music will likely be more captivating in Udio’s world.
Of course, this is just a guess, as the developers will likely never reveal this, but I feel that Suno was trained with more „hits”, catchy songs, while Udio was trained more on indie titles (although I note that one generated song had a very „Michael Jackson like” singing voice).
It might be a marginal need, but mixed-style songs are important to me. So far, Suno “plays better” in this regard, although after many attempts at prompting, Udio can also be persuaded to simpler style mixes (note: more than a week has passed between writing the first letters of the article and its publication, and during this time, Udio seems to have improved in this area).
Now, I deliberately brought an example of a style mix that both platforms successfully resolved. In this example, the goal was to mix gospel and hip-hop, which they solved quite similarly, although it’s clear that they draw from completely different eras and sub-genres under these style names.
Digital Grace – Suno
Digital Grace – Udio
Languages
We now take for granted that in English, TTS also brings fantastic results, and this is the case for both Suno and Udio when generating singing.
In addition to this, the developers place a strong emphasis on other world languages, such as Chinese, French, or German. Smaller countries or languages used by very few users are given less emphasis (which is perfectly understandable from a resource management perspective). I also tested the applications with two languages I’m somewhat familiar with but less commonly used, which in our example will be Hungarian (my mother tongue) and Icelandic (which I’m currently learning).
For Hungarian, I generated a sort of “ad jingle” for a friend’s bike service (BTW if you are biking around Pécs call them in case of technical problem) – Style Phonk/Hip-hop. Let’s see the experiences:
Suno:
The musical base perfectly matched the prompt
The Hungarian pronunciation is acceptable from the 3rd line.
Prompting phonetically yields quite a good result.
Note: Since I’ve been using Suno in foreign languages for months, there’s a quirk. In the first 1-2 lines of the songs, it use an English pronunciation, even if you include the target language in the text or style prompt.
Udio:
I wasn’t entirely satisfied with the musical base (it automatically completed the prompt – this can be turned off)
The Hungarian pronunciation is generally acceptable
Prompting phonetically yields variable results. Sometimes excellent, other times more amusing
Note: If it can’t interpret part of the text (yes, it interprets, as it gives a random English title according ot he text), it fills the relevant parts with completely irrelevant gibberish.
For lovers of Scandinavian folk music, I generated the lyrics of an Icelandic folk song with both applications, with a broad style definition (experimental electronic, nordic), resulting in similar style outcomes, and here comes the question of taste. While I think Suno performed better with the Hungarian phonk/hip-hop, in the case of Icelandic experimental, I judge Udio to be much better. Due to my weaker language skills, I wouldn’t make a more detailed analysis of the pronunciation. I hear problems here and there, so the “AI pronunciation” is still easily identifiable, but I didn’t find it offensively bad or outstandingly good in either case.
Have composers and music producers really lose their jobs?
Yes and no!
I believe that those who did low-level or mediocre work, or even “mass-produced” and can’t use AI to improve their own work, are indeed doomed. Those who either create outstanding works and/or can utilize the possibilities of AI can be greatly helped.
Although I’m no longer actively producing, I asked around a bit. Those who are open to using AI generally use it in two ways.
First, for brainstorming, because everyone has bad periods, and even in a good period, a good idea, further developed, can become fantastic. Plus, AI doesn’t „think” like humans, so you can pick up nuances that wouldn’t have occurred to someone at a given moment.
The second most common practical application was replacing sampling. One producer I asked primarily makes hip-hop, while the other loves to make electro-swing. In both styles, it’s crucial to have good samples available. And for them, a good AI-generated song can replace a lot of “research work”, subsequent “tweaking” and cutting.
Is there really a new king of generative music-making?
My answer to this is a clear no. Rather, I would say that a kind of dualist system has emerged regarding music AIs. Each has its own weaknesses and strengths, which they seem to know well, as they strive to develop in the gaps left by the other. If I had to draw a single conclusion from the entire testing phase and this article, it would be the following: With the arrival of the new „big gun”, a two-player market has formed, and the competition between them will be very good for us, users!
+1 Which One Should I Use?
Unfortunately, we won’t be able to make this decision for you. You’ve seen that each has its own childhood diseases and strengths, so it largely depends on what you want to use the generative music for. If you want to put music under your reels videos, which almost everyone watches on mobile devices, then maybe you can quickly achieve a suitable result with Suno. If you want background music for a YouTube video, then it doesn’t matter which platform you use. If music plays a prominent role in your project and sound quality matters, then Udio. And of course, there can be thousands of variables in determining what kind of audio material you need. Only you will see this.
If it’s a reference for you, then in my case, I’ll proceed as follows (for now):
Since I use music AIs quite a bit even in “marginal” languages and audio quality is rarely of paramount importance, I’ll probably continue to use Suno in 70% of my projects, but Udio has already been incorporated into my workflow for the remaining 30%.
Have a creative day!
About the author:
Active Marketer, AI Freak. retired music producer. From 1998 to 2015, worked as a music producer and also ran a small music label for nearly 5 years in the early 2000s.