Imagine selling A.i music as the future of music to investors, who agreed to drop 100's of millions training on the( unauthorised) data of real musician's and calling your business ethical. Oh and don't forget putting Beta releases behind a paywall since day one.
Out of pure curiosity, I made a solid album in a month 3000 credits later using old demos, but can confirm the tech is so horribly hit and miss and v5's language model is actually worse at pronunciation than 4.5.
Suno can help realise a rough demo to a mock up of a full production, but please guys don't waste credits on Suno's stem separation, use UVR and you'll get a way more usable set of stems. You can then delete all the whooshy washed out noise and build out a real track and Re-record the parts so it won't sound like a machine mimicking a human performance.

 

Last edited: Oct 5, 2025 at 5:24 AM

Every AI is just as good as you prompt it. As much as I'm not a fan of people now just posting AI music without disclosing it I like to play around with it for ideas or generate vocal tracks to start one of my own songs and you can simply prompt SUNO with BPM and Key and it will usually work I have none of these described issues even with V5.

 

Such as? Please do tell us all even just 1 prompt that will deliver a pristine quality mastering studio level result.

This is just not true in my extensive tests. I scoured the internet for these magical unicorn prompts and used several ai chats to get suggestions and tried them all with a mix of unusably horrible and iffy results in almost all instances. Im not saying it is impossible to get some decent sounding parts and ideas because it does sometimes spit out a great idea, but I am saying it is so random and inconsistent and even when it is good when you separate stems and midi it all falls apart immediately. I spent thousands of credits and countless time but I have still not gotten a perfectly constructed, well balanced and stable mixed and mastered song from any prompt. There are always problems to solve and usually more than worth wasting time on. And do this all for what is a song and tracks with unremovable watermarking and you dont own the full rights to? When you say you dont have any of these issues its unbelievable and makes me think you work for Suno or something.

But you can shut me up by proving it? its not like the song is yours or anything so you have nothing to lose by posting it here right? Still waiting.... I actually really hope someone can do this so I wont feel so bad about all the time and money wasted.

 

I agree, I have also tested Suno pro extensively and it takes a lot of fails to get something close to a good take.
Prompts will usually be more accurate if you feed the style bot prompts with specific technical musical terminology. " Tempo, time sig, mood, expression, feel, sonic sound signature, appergio, legato or staccato, vocal tone & timbre, rhythmic patterns, and what production or mix style it's inspired by as it won't allow referencing artists or songs by name.
Suno doesn't even have a proper tutorial in it's knowledge base, because they expect you to keep guessing and feed the credit cookie monster. It's a fun tool for jump starting demo ideas but a serious musical counterpart it isn't, Yet...

 

Last edited: Oct 5, 2025 at 7:12 AM

Beware of a.i sycophancy. I caught chat lying to me about feedback on some lyrics and it apologised but refused to admit it was being dishonest. Just like a dating site, it's been programmed to feed off my approval rating , to build a fake relationship to harvest more data eggs.

 

They call it Glossing. I like how you can ask it which is a better audio software product and it will refuse to say which is "better", but then if you ask it to do a shootout like product #1 vs product #2, it will not have any qualms to eviscerate the loser. Or it will try to snow you, but then if you ask "so product one really sucks?" and it will answer "Exactly!".

 

There's something crazy going on with how these have been programmed, but then it has to take data from so many sources...

When it comes to feedback its pretty similar to what you'd get on this forum, like cut mud, enhance presence and it can gather that info from many sources, including all the nonsense like 'its your track, you can do whatever you feel like', which isn't helpful in making a track that sounds typical, radio tracks, album tracks that you typically buy, or probably the best reference point - could you get a distributor to put your track in a store (without having to pay them, because they think it will sell)... I think at that level you've done something decent.

By contrast to Suno, where it seems you don't have rights, GPT is actually usable - except when it crashes (frequent or regular). I've taken a track to a balanced finish in 2 days of feedback sessions. 3-4 revisions... That's faster than Cubeadooby in getting to a decent finish ;)

 

Your looking at it wrong. Its not about creating a fully mastered song. Well, maybe for some it is, but not for me. Its rather about using it as an idea generator, or for sample fodder. You can get good quality results regardless, it just takes some finessing.

As far as the stems go, there is now Suno Studio, which allows you to generate individual elements, one at a time. No more bleed, no more wonky stems.

Personally, i have not had major issues regarding bpm drift, but do occasionally encounter an issue. Luckily its fast, and easy to fix any timing issues in ableton, if i do plan to actually use the stems within a song, which is rare.

 

Yes its about all it is useful for right now just to generate ideas that you must recreate from scratch yourself in your daw or hardware. But in my opinion this defeats the whole purpose of Suno and especially paying for it! The best quote from the thread @marshall stack 666 "keep guessing and feed the credit cookie monster"


I dont use it for full mastered song I am just saying the results I get from prompts generating songs, then separating it into stems in studio, then downloading audio and midi stems from there and put inside daw and I use Ableton too. The results are consistently bad 99% of the time I try it. The songs fully generated are atleast cohesive, but separated and dropped into daw as auido and midi stems it all falls apart even with ableton warp on. In addition to the other problems I mentioned there is always a "whooshy washed out noise" which @marshall stack 666 best describes it.

Then add the Suno watermarking on every thing generated it makes no sense to pay for this. I really wonder how many people still dont know about this watermarking they do. Watermarking is in EVERY single sound it generates that can never be removed! I see people saying stuff like "oh i just use it for random samples here and there that I twist and chop and reprocess so it is unrecognizable and original" and what I think they need to know is none of that matters! No matter how much editing and processing you apply the watermarking is still there in your song now and that song is no longer yours all because of even just one small micro second snippet of sample, now it is watermarked with Suno and will be red flagged removed from streaming apps and especially if you didnt "feed the credit cookie monster" for generating the original sample you dont have clearance to even use it for more than "personal use"!

This also brings up the interesting point about sample packs because many are now mass ai generated so if you use sample packs this watermarking also unknowingly applies to those you are using in "your" songs. And if you use cracked samples youre really f+^%$#*d!

And also I think it is very dishonest how they advertise the pro/premium packages because it doesnt tell you that even though you paid more money for commercial copyright use all the files still contain Suno watermarking that cannot be removed! Just because Suno says you can use it doesnt mean streaming apps will allow you to and in fact most now strictly prohibit it! If I knew this before I would never have paid for the credits. Very shady tactics I think.

Yes as i mentioned I also tried uploading my own parts many times and asked for single track creation but it was worse in most cases. Yes the bleed from other instruments is gone but the sound quality is still not pristine professional studio song ready. For example I will upload simple 4x4 kick drum loop of 8 bars. The sample is high quality from analog source and fully processed that I have used on a released record and know works in clubs eg perfect for my uses. I ask it to create a bassline eg (random from the top of my head) "Techno bassline, deep and hypnotic, 130 bpm, rolling bass" and the results are laughable not only because of personal tastes such as I dont like the chosen notes or pattern or timing which is the first failure part but because the sound quality is just so bad. If I take that single audio track and extract the midi and put inside daw to use with my own synth it is not the same as the first suno studio generated audio file the midi and audio notes and timing dont match!

I will say I find the most consistent generated quality audio is surprisingly vocals. But anyone who uses a vocal from Suno even if your own lyrics is begging for their music to be banned everywhere and their artist or studio reputation completely destroyed because of watermarking. It is not if but when it will happen to you if using Suno vocals. so again I say why even bother wasting so much time and money when you can hire studio session musician for a singer and pay the same or maybe even less money for better quality delivered when and how you want which you will 100% own forever and never be banned for?!

I dont have the answers for this but can only say they have a long way to go and should remove these f*&^%*&^ watermarks, ALL OF THEM! if they ever want my cookie jar money again.

 

You are completely wrong. UVR CANNOT seperate Ai music stems the way Suno can. I've been through hundreds of songs and trying both. Yes, Suno's stem separation is not perfect, but it's WAAYYYYY beter than UVR.

 

you can upload audio. So just turn your midi into basic audio a simple sine wave patch or grand piano. then use that to guide your song.

 

Youre thinking about it making music like humans make it. Its a whole different process. If you want to build a track like humans do you can do that with any daw. Thats the power of things like suno. It goes about it in a differnt way. A way that can create cool new thing if you know how to push it.

If you want AI to build a song like humans use Ableton MCP

or a platform like soundful.
https://soundful.com/


If you want something totally new i emplore you to use suno, udio, sonauto, producer.ai etc. to give you things you wouldn't do in conventional ways

 

Gotta agree with @curtified and @shinyzen : Suno is not able to make a human sounding song. The point is not that. It's a sound database where you can generate the style of sound you're searching to your human production so you can save some time and money on sample kits. If you want to make a "human sounding song" you gotta learn how to use DAWs and playing real music instruments, and i believe it will be as this for a long time. You all remember the case of the AI generated folk rock band on Spotify. It sounded awful. So, this is just the point i'm bringing to the table when we discuss about this tech.

And putting raw stems here or full AI songs will not bring anything to debate for me. Anyone can go there and do it and see if it works for you or not. And i believe most of the folks here know how they sound.

 

fully agree. Taste is what filters out the slop. AI or human made music

 

it isnt through "watermarking" its just the way AI music is. all the top end is chopped off. Its generating thousands of spectrograms through diffusion. Today its easy to detect. So yes if u go to any AI gen platform and say "make me ____ song" and try to upload that to any of the AI detectors it will most likey flag it. It helps so Spotify isnt flooded with AI music making it so actual musicians music is lost in the slop.

as for this:

Again its false. no watermarks on the audio. But any manipulation of the audio or adding any sort of actual instrumentation from an outside source will help not trigger the your music is generic AI music detectors.

https://www.jenner.com/en/news-insi...sh-copyright-issues-for-lawyers-daily-journal

that is the current laws setup to help with AI slop music and generic Music spamming to DSPs

You can also get around what you call "watermarking" by uploading your own work and building on that with a AI tool.

 

Last edited: Oct 7, 2025 at 12:15 AM

Id love for @Rolfy to show me these watermarks they are talking about on a random suno generation. Who is auditing these said watermarks?

 

Last edited: Oct 7, 2025 at 12:46 AM

Wait. what?

Preface;
I completely detest these round robin vegas style slot-machine music "creation" traps. They are all without a doubt as completely far away from actual "AI", distance wise, as planet Pluto is from the star SOL... The "artificial" part being the ONLY thing applicably correct to be taken from the acronym. But "intelligence"? Please... The only thing even half-assed "intelligent" about these garbage bins is the venture capitalist entrepreneurial developers trapping unwitting pawns into forking over actual $$$ for such nonsense. But, I digress...

The question; WHY the hell would one want to "create" a "song" with a non-constant BPM? pffft.. Sure - tempo changes. I get that but a constantly random BPM? really

 

Because there are infinite ways to create audio. This can be one of them. Just because you detest a way someone gets wav's for their timeline doesnt mean someone else cant get a waveform this way.

I see it on here all the time about DAW's or audio tools. Its like peoples identity gets threatened by the way another person makes music. Its all just waveforms at the end of the day. How does one persons way of creating hurt you from making things your way?

 

Well, i don't know if Suno is watermarked or not, but the samples you download on Splice for sure are, plus, it's a much more risky way for a copyright strike using royalty free samples than any AI generation. You're not the single one that loved dat KSHMR EDM loop, i can asure you that. I know a few cases of people that got their jobs rejected by labels for using famous 3rd party samples.