i was talking about a "mix feedback" thing, not about generative music....

 

I didn't realise GPT can already give feedback. It told me its capable so I'll try that today and see...

But generally if its as bad as this Suno post suggests we're some way off having something usable... I'm looking forward to seeing what GPT can do already...

 

That's the problem, anyway I personally know artists that uses it as a "composer assistant" (for background music mainly) and are pretty satisfied with.

So it depends on your needs, IMHO.

 

It sounded really corny. Unusable i'd say. But, at the same time they got a sound enhancement, some of the features seem somehow not to do what they're supposed to do. For instance, the "extent" function is not working properly for me since v5 and some of the results, if i manage to extend at least part of the sound, don't have nothing to do with the actual sound clip i uploaded. If i change to 4.5 will be the same thing. I'd say it's not only a problem with v5, but with the overall implementation. If i use prompts only, not audio, it will work, but some of the results sound glitchy and nonsense at v5. At the older versions, the prompts sound ok. But i don't like texting prompts, and generally saying, i don't like the results if i don't feed the AI with audio i produced myself. So, the way it is rn, is unusable for my purposes.

 

Last edited: Oct 2, 2025

Yep. And with moores law things are just going to get better faster. Back when suno was on discord only i had to generate 100 songs to get 1 good one.

then v2. cut that down to 75... v3,4,5...... makes things easier and better. Exciting to know how to hack and use the tool today to help me in the improvements for tomorrow

 

Just use the older 3,5 version of suno guys bad mix but great results

 

Yeah you're not wrong... Cheers for this, I had no idea GPT could do this already... Been doing a feedback session today, like where it might take weeks to get feedback, go away, fix a bit etc... This is instant and has actually been useful... Its some way off being brilliant but its useful already.

Thanks.

 

awesome. glad you have found some use of it! I think soon we will have a dedicated app / agent that can "listen" to and analyze our material even better, and give better notes. Its also important to remember there is no such thing as perfect, so don't rely too heavily on feedback, trust your gut and your ears. Something my therapist told me which stuck with me, "Done is better than perfect". Take some feedback, make some tweaks, do another round if needed, but don't get stuck in the endless loop of feedback and tweaks.

 

Yeah so very true. The best tracks are ropey as fuck, it adds something special... What this is doing though is incredible. Its rough guides and 'someone' to chat to real time and ask 'which should I try first' when looking at all the dials. It has so much access to knowledge at this point it can give some pretty good guidance... I just put in my dry vocal and it did an analysis, printed the average and suggested settings based on what sounds shit in my voice etc... Like you say only guidance, but I'd have no idea what to tweak otherwise, I'm shit at decision making... Absolute 'game changer', I fuckin hate that phrase at this point, but it genuinely has removed years of guesswork and given me hope of getting a polished finish... We've done 3 feedbacks today already, making a load of moves it suggested and its incredible really... Even better than a forum etc, where people tend to be kind. It doesn't care and just tells you what to change... Incredible really.

But yeah, great advice actually, just get a good mix to the point where it stops criticizing and call it done. It pretty much already guided me most of the way there... The guide vocal needs redoing, but its helping me to set up my Apollo for real time compression etc, then suggest some 80s sounds for Synthwave... Doesn't really matter about being amazing, just good enough is good for me.

Can't believe we're already at this point though. I wouldn't have thought to try it, I thought it would be a long way off still.

 

It can't be that stupid - you're promting it wrong.
Joke aside - it will be better, it just needs more data.
Compose a banger, upload it to spotify and wait till the bot comes 'round.
Then, after a year and a day, suno is gonna be awesome. You'll see.
Or that whole AI bs is a multi trillion bucks scam.

 

On a side note, the OP is... ahem... "no longer with us"

 

Suno is good, as long as you know how to use it.

 

I have a theory they they are with us often.. they just posted the orig post anon to ruffle feathers on here or to not show who they truly are.

 

Facts! TBH thats any tool to create with.

People can create slop with a daw too

 

How did that happen? I thought we were quite nice. Within our limited capabilities that is

 

Suno can be handy for developing those 8 bar loops into something more. Sometimes it will come up with a vocal rhythm/phrasing that fits nicely and it's really good at bridges and pre-choruses where you need a little bit of contrast to the rest of the song.

I don't use any of the audio it generates but I have duplicated a few chord sequences and melody snippets here and there. Oh, and I like to give it a stupid topic to sing about to discourage me from nicking the words.

Yep, that one's about bananas.

 

I agree, something similar. Probably already banned. He went bananas with the swearing, holy shit

We were nice.
As opposed to... him

 

Hmm I dont think he was completely wrong about what he said and some of these statements make me wonder how many people posting comments here work for Suno or get some affiliate commissions?

For myself I have tried Suno 4 and 5 and all I can say is it isnt a serious studio tool and the only way it ever will become one is if it flips its content generation to recreate what humans do in a studio because right now its backwards in my opinion. Im saying instead of generating a song then stem separating and then audio to midi conversion it should start with midi to create the audio and then using those parts generate a mixed master of the song final generated. If not it will always have bleed from neighboring audio frequencies in the audio stems that are clearly audible in my testing and then converted to midi it always results in a mix of random out of tune and timing midi notes. iin my tests I found that all of the audio stems didnt have the full instrument parts throughout the entire song and would drastically change over time sometimes completely changing into different sounds from the start and end of the songs along with the bpms drifting. They are often mislabeled and the tempo and keys mismarked. I think its a real major problem for Suno and other ai apps to overcome. When I also uploaded my own music and parts it rarely ever gave me the accurate results no matter what prompt I use and I tried hundreds of supposed expert recommended prompts even for prompting simple 2 part stems it failed no matter what prompt I used many of which werent my own prompts but those suggested by "experts" and ai itself

Its also important to tell people about all of these files are watermarked and you can never escape that without recording digital to analog back to digital and using specialized a/v and software data restoration processing to remove it but will also severely damage the files sound quality in the process making it all a pointless and convoluted exercise of time wasting. Suno will always know if you did or didnt pay for the file and streaming apps will remove your music because of it and its very probable in the near future it will be completely banned everywhere so all of those songs containing any ai parts will be removed.

Testing and morality aside I am not against to trying it again later but my conclusion right now is im not impressed and I will never pay to ai generate songs and parts that I dont own because I take this seriously enough and whats the point. If I cant use my mind and skills to create something I can pay real human studio musicians work for hire fees or ghost producers to get pristine high quality sound and midi and own all rights without watermarks 100% forever. Unless you dont care about the rights ownership or sound quality and just use it for disposable tiktok shorts or something and have lots of time to burn fixing endless audio and midi problems and dont care about your stuff being banned from streaming apps, then until Suno can offer the same as real musicians and studios I think everyone is wasting time and money on this.

Does suno suck? I think it doesnt always "suck" in ideas like maybe 1 or 2 out of 50 prompt piqued my ears and even the threadstarter admits this but lets be real it is galaxies away from being anything amazing as some biased people are shilling here.

If I am wrong then please prove it to me and everyone else. Im not trying to offend you by expressing my opinions and sharing the truth about all of this but why not post your amazing ai songs and audio midi stems to show us all how great it is to put an end to this debate? ....... Ill wait

and before anyone asks we dont need to post the same proving it is not amazing quality because we all can try for free and quickly find out ourselves it is a true reality.

Peaaace!

 

@curtified works for Suno. The rest of us is just discussing (i think).

That was a feature i always missed on Suno. The lack of possibility of uploading midi to feed the input is kinda frustrating for me as well.