ao link
Affino
Remember Login
Remember Login
Remember Login

AI-Generated Music is incredibly already coming of age!

AIAI Song GenerationAI-Generated MusicResourceSunoUdio+-
TweetFacebookLinkedIn
AI-Generated Music - Suno and Udio
Sharelines

Suno and Udio in Arms Race for best AI-Generated Music Platform

AI-Generated Music is really coming of age

AI-Generated Music delivers incredible levels of sophistication already

So I’ve been spending the last couple of weeks exploring the two current leading AI-Generated Music / Song Platforms - suno.com and www.udio.com.

 

Chat GPT’s Suno has been around for a while - it’s V3 currently I believe, while Udio - created by 4 former Google DeepMind staffers is currently in Beta and has only been live since April 10th this year.

 

Each platform is incredibly sophisticated as it generate whole songs - instrumentation, vocals and lyrics wholly automatically - you just need to be clever with you prompts - to ellict the appropriate response. It’s still something of a lottery, while both platforms are capable of incredible results.

 

I look at Suno being more like Dall-e, while Udio reminds me more of MidJourney. Suno is typically more accurate to your prompt / description - while Udio, like MidJourney - is capable of some really creative outcomes.

 

Both can be a little loose with ’Genres’ - it’s a particular current weakness for Udio - where you often get quite the wrong genre to what you’ve prompted, Also key commands like ’female vocals’ are often ignored - and you get something in a male voicing instead - I found that to be very much the case for ’Metal’ genres, including ’Rap Metal’ - in fact I gave up on my Rap Metal creation after 12 attempts on each platform - as they just weren’t adequately representing the prompt. So there is still plenty of distance to cover. 

 

Also I feel that the next phase of AI-Generated Music is going to allow for more Editing and Re-Shaping / Remixing - essentially covering off more of what something like Hit ’n’ Mix’s Rip X DAW Pro allows you to do - edit out the various stems and wholly re-master the tracks - bar by bar!


Suno.com

Suno.com Generative Music AI

Both platforms work on a similar premise.

 

You enter a prompt which defines what kind of song you're looking to generate - Genre, Instrumentation, Nature of Vocals, and Song Title.

 

Suno is generally quicker to process prompts than Udio.

 

Both platforms generate 2 versions - giving you 1/2 / A/B choice options.

 

You then select which version you prefer to further extend and / or remix that stem.

 

A key different is that Suno generates a full circa 2 Minute track (1:52 in my case) with intro and outro already tagged on, while Udio generates 33 second segments - which you then extend in 33s segments - and need to add specifically Intro and Outro extensions to complete the track. So in my main experiment here Suno generated the perfect track 'Butterfly Dreams' at first attempt, while it took me a dozen attempts on Udio to get something similar. There is a huge degree of randomness as to the suitability of the generated music - in relation to the prompt. The sound quality is always exceptional though.

Each platform has its strengths for sure - Suno generally has a more accurate prompt response, while Udio often generates more interesting vocal and instrumental textures.

Here below is the very first track I generated on Suno - Butterfly Dream!

The prompt was as follows :

"ilting female vocals liquid drum and bass sliding bass song about butterflies"

Butterfly Dreams


Udio.com

Udio.com Generative Music AI

Udio at its core operates very similarly to Suno - it generates Instrumentation, Vocals and Lyrics from a prompt - and presents you with 2 examples each time.

 

However the size of segment here is just 33 seconds, and you then build up the track by adding 'Extensions' which can come before or after he root stem, and you can then select to add Intro and Outro 33 second segments.

 

In some ways I prefer the Udio interface - it's more visible and obvious in many ways, while Suno has some tricks of its own - including the excellent and very handy 'Instrumental' selection button.

 

For my Udio-generated track I had to add 3 further 'Extensions' to the track to bring it up to the 2 minute mark - or specifically 2:10.

 

This track 'Kaleidoscope Heartbeats' took me more than a dozen attempts to generate. Udio kept generating music in wholly the wrong genre, and with male vocals - so multiple attempts were necessary.

I often consider that experience similar to being at the opticians - where you're constantly presented with A/B choices and you have to decide which option suits you sensibilities better!

So often it's a case of perseverance!

There are of course some glitches to be found on each platform - I've been wholly unable to generate a Video download from Udio - it just flashes up the message "Please try again later. Video is being generated!" while nothing ever happens there - you leave if for several minutes or hours even, and still get the same alert flashing up.

By the same token I was wholly unable to Extend by Suno track - it made several attempts - but each time just came up with dummy files - which had 'Credits Refunded' labels applied to the empty files. Actually - note that Suno works on a system of Credits, while Udio works on a system of 'Daily Priority Generations'.

Generally Suno is quicker, but Udio can be more creative, and is more generous with the attempts it affords you.

In any case here is my Udio generated track 'Kaleidoscope Heartbeats' - which was generated with the same prompt as 'Butterfly Dreams' but took a lot more attempts, and includes 3 Extensions.

Kaleidoscope Beats

Kaleidoscope20Heartbeats20ext20v1.1.2play_arrow02:10

As mentioned - despite numerous attempts across numerous tracks I was never able to download a Video version on Udio, while that worked instantly for me on Suno.

On the other hand - Udio extended tracks with ease, which Suno never managed to do - just kept generating empty files with 'Credits Refunded' tags.

 

The format I've spent more time with us Udio - as it allows me to shape my track better, while both platforms are equally capable of incredible and very creative results. Obviously the lyrics can be a little cheesy and awkward at times - but all-round this is incredibly impressive.

 

My very first AI-Generated track was the below Udio-Generated 'Jetstream Adventures' - which I was able to extend a maximum 7 times - to produce a pretty impressive track length of 4:22 - which does seem to be the ceiling currently for that platform. I kind of ran out of road on that one - so it doesn't have an Outro - but still kind of mostly works fine!

Jetstream Adventures

Jet20Stream20Adventures20ext20v2.2.2.2.1.1.2play_arrow04:22

So far this has all been very much on the Liquid Drum and Bass tip - while I've experimented with various genres - including Jazz, Metal and Disco too!

 

As a final sound sample here follows the Clarinet, Piano, Bass, Drums Jazz Swing Quartet song I generated for my Dad - again courtesy of Udio!

Sunbeam Serenade

Sunbeam20Serenade20ext20v1.1.1.2.1play_arrow03:15

This took quite a few attempts to get right - as Udio couldn't get the Genre quite right to being with, and threw in a whole load more instrumentation than it was supposed to at times.

In any case I persevered and created a proper Jazz Swing track with mostly the instrumentation it was intended to have - even though the Clarinet sounds a little Saxophone-y in the beginning.

 

This is all very typical for both platforms - both were abjectly poor in attempting to generate female led vocals Rap Metal - mostly they just totally failed - and each version was very much male-fronted!

 

You need patience and perseverance to get the best out of both platforms - which I still feel are very much in their infancy. I can only imagine how good these will be in a year to two's time. With the ability to edit Stems too in all kinds of incredible ways - like having a proper collaborative music creation session with a a band - interacting to your commands.

At the moment it's a touch a lottery at times / crapshoot - while you get incredible results a lot of the time.

As a next step I will be a little more active in creating my own lyrics, and seeing how the two platforms deal with those. You sometimes get odd vocalisations - trying to fit word into the meter - where both platforms occasiionally 'fudge' things - a little like the vocalist forgetting the lyrics and mumbling at times.

Generally though - this technology is already totally amazing!


Suno vs Udio

Suno vs Udio

I selected Suno.com and Udio.com as they're generally held to be the leading lights for AI-Generated Music Platforms. And each is capable of delivering incredible results - but not without some challenges and glitches.

 

The Genre association can be particularly flakey at times - while Suno is typically more accurate in accordance with your prompt, albeit in a MidJourney kind of way - Udio can be more 'creative'.

 

I feel the best track generated was the Suno-created 'Butterfly Dreams' while that was only 1:52 minutes long, and I was wholly unable to extend it. Suno tough did deliver a Video Download pretty much instantly Generally Suno seems quicker at processing your prompts. Where you can on occasion wait several minutes on Udio - but Suno is typically just a minute or so.

 

Generally I prefer the Udio interface and help references - in some ways it works a little slicker - while the output of both platforms is pretty much equal in quality, I was unable to Extend Udio tracks beyond 7 Extensions - which of course was a lot better than the 0 Extensions / Credits Refunded nonsense from Suno. While Udio seems wholly incapable of generating video downloads - which Suno seems to do rather effortlessly.

 

The music industry has of course reacted in a typical luddite fashion - calling for both platforms to be banned. I'm sorry - the genie is out of the bottle already!

I still favour proper Live Music Gigs - while I also loved the very much AI-Assisted ABBA Voyage Show - which still had a sizeable live band on stage at all times.

There are still magical songsmiths out there in the world for whom these platforms aren't really a threat right now - but they probably will be - with a little more sophistication - in two years time.

I don't really want to go to much into the ethics of Generative AI Music - this article is really just about the technology and possibilities.

I feel that a good way to end this is with proper real live music - and showcase the latest brilliant single from our good friend Steve Hinds' Motorcycle Display Team Band. Steve is of course Head of Digital at Travel Trade Gazette. This song deserves to go all the way to the top!


To The Endless

close
Stefan Karlsson
Posted by Stefan Karlsson
TweetFacebookLinkedIn
Add New Comment
You must be logged in to comment.

Did you find this content useful?

Thank you for your input

Thank you for your feedback

Blog Navigation
Blog Navigation

Upcoming and Former Events

Affino Innovation Briefing 2024

PPA Independent Publisher Conference and Awards 2023

Driving business at some of the world's most forward thinking companies

Our Chosen Charity

Humanity Direct
This site uses cookies, You can manage your preferences by clicking cookie settings, or simply accept to gain the full experience.
Cookie Settings
News & Insight Assistant

Login or Register to get access to Affino's services including the AI's Insights, Support and much more.

LoginRegister