📚┃suno-school

[cunning mulch]

I know one that had at least 2, and possibly could've been 3... might be worth saving off

although the 'third' could also be a 'filtered' version of one

[pine dagger]

so 2 voices is not easy for the AI so i think more voices is hard but you can try it

with different Prompts

[cunning mulch]

I'm talking a song that's already been generated

[pine dagger]

make a Cover you mean

[cunning mulch]

no

I mean, I'm considering making a Persona from that song

[pine dagger]

oh i didnt work with persona actualy sry im not the right guy for that, sry

[remote rivet]

Feature request: I would like Personas to be able to do covers. Thank you.

[polar spade]

A lot of words about nothing really, as its likely just, but.... I do notice the times I do a 70s like song style, a Jackson funk, yacht rock, jazz rock, mellow gold, etc. it's been those that will give me a fade about 3%. The style in the 50s-80s. Even some Progressive Rock fade-ins as well. Whereas other styles never ever have, as I recall.

I'm one to try to strategically put this concept into the styles prompts, adding whatever you want in metatags, to see if it's got just enough weight at the end of a list, to influence anything.

I use opera for trying to get some warm solos and duets since opera uses individual singers and a choral. Barbershop for harmonies, ensemble sounds and duets for similar reasons....

I DID get a fade in the middle... of a glitch vocaloid sounding song with the style:
"Hypergagic experimental jazz pop, glitch pop, sycopated flutter" (hypergagic means nothing, as I misspelled it, should be hypnagogic) but both 'hypnagogic' and 'jazz pop' are styles I feel hark back to fusion and 80s era, so maybe, just maybe that's why.

[split escarp]

I'm getting a moderation error. Using lyrics for a public domain song and being told its not allowed.

Tried with another public domain song, and that also gets an error

3rd song now, all public domain

[untold briar]

Just bad luck, unfortunately. The copyright databases have a lot of false positives. It's gotten better than it used to be, but it still happens from time to time.

If it's happening for 3 songs in a row, that is odd though.

[brittle apex]

I've run into the same, try flexing the spelling a bit in ways that won't break the pronunciation. eg away -> ehway

There is no reliable way I have found

[sacred cobalt]

how many billion parameters does the latest SUNO AI model have?

and what is its size in gigabytes?

[charred merlin]

🤷‍♂️

[pine dagger]

All songs i make on Suno, text are Write 100% from me, so you cant put Artist like Kontra K or anyone in Music Style too

[pine dagger]

I found it, in Music Style put in "BothGender" and when you have a own text than say for example [Hook | Girl-Name] and in Verse [Verse | Boy Name] so its possible that the song make two different persons Womand and Man

[worn flicker]

There is a reliable way to make it happen, but not necessarily in the order that you want it, so you would need to make sure your lyrics are not saying him or her if you use pure Suno. The other day, a person I forgot who already so I am sorry for that, but they showed me this site.
https://audimee.com/create
I haven't actually given it a try yet, but it offers free stuff so why not try it if someone really wants the male and female to be speaking in a gendered way.

[sacred cobalt]

and where can I see what public persones already exist? (and see the tracks that were created from them)
https://suno.com/blog/personas

[brittle apex]

The persona I created Jack&Jill reliably makes male and female duets, but I have not found a good way to control precisely which voice sings when

[polar spade]

Just to point out. The Timbaland lyrics say "her". My songs are split between male and female vocals, all singing the "her" line. One ends in a duet singer HER. The whole song sounds like a love triangle. But then it's not the Initial lyrics I start with. I'm covering or extending Mr Mosleys original. So, there is that.

[worn flicker]

I find that both Female Vocals, Male Vocals, Duet in style makes it pretty consistent, the the tags in the song makes it swap. The only issue is that it likes to reverse the order regardless of what tag. However which ever is singing that section, if you put (....) as the last line of that section then the other vocal will most of the time sing it. From there you somewhat defined that (...) is going to mean that other vocal.

Meaning later like
[Chorus] [Duet]
(....), .....,
......, (...)

Will swap mid verse

[pine dagger]

i type a text above Sysmai there is a Way so for me works it excelent

[worn flicker]

I couldn't find the original so I chained it...sorry...lol

[pine dagger]

yes i do it so

Sry text is on german
[Intro | Anka]
Ich denk nur an dich, so viel an dich,
Doch der Schmerz bleibt, ich fühl mich nicht leicht.

[Part 1 | Manuel]
Du gibst mir das Gefühl, dass ich denk, dass du mich liebst,
Ich versuche es dir zu zeigen, aber ob du’s merkst, weiß ich nicht.

But the names are Important and the Tag "BothGender" so it works perfect

[worn flicker]

Yep that works too

I have seen the names defined and used as well, I can't say I haven't seen it swap though

[pine dagger]

but i see you need in best, international namens so English namens few German names are not working ^^

[brittle apex]

I did exactly that, but it isn't fool proof. I think that would work most the time if you are swapping verses but switching multiple times in the chorus with unison singing, the AI had a hard time keeping it straight. https://suno.com/song/67b0707a-50d7-4b5c-bee3-b59b0113384d

[charred merlin]

They're available on user profiles if that user has made them public

[worn flicker]

I don't normally put my tags together though, it seemed to skip the instructions when I did that, but I never tried the | since that works in Linux I don't see why it wouldn't work on here. I was trying with commas

[pine dagger]

yea its not easy so you kill credits any way when you would like the best song ^^

[worn flicker]

That site I posted is supposed to work well.

I just wish I remembered who gave me the listing

[pine dagger]

"Hip-Hop, German Rap, Love, BothGender"

I think its important to put the last Tag as Main tag so otherwise he killed any tag xD

[worn flicker]

I love how many different ways work in different ways...haha.
The fact BothGender crammed together works is funny.

[pine dagger]

so at this time it works at the first try directly for me, i dont know is that luck or its work realy perfect xD

[worn flicker]

I mean who figures this out other than a typo...lol

[polar spade]

I asks suno to write a duet. Ironically, though Suno always tags it's own songs in this manor, it randomly sings them any way it wants. This quick test I just did, all three verses are close to an ABBA sound or girls choir.

[worn flicker]

I know the way I do it is very consistent, but the verse swap thing is a problem. Other than that I have many duets. I also must say that your style you choose also helps. If you list a style that tends to have duets in it, you are more likely to get it. So Broadway works quite well

Even if you are doing Broadway. You put it later in the style prompt

[warped furnace]

What's up with Suno? Can't he understand genres anymore? Have generated a lot of Latin, but just now, it gives me everything but Latin genres. Wanted Merengue and get some operatic Rock. 🤣

[worn flicker]

It has been taking Latin and making really deep choral Vocals it is weird

[pine dagger]

I had hard work so a German Text, with two different voices, me and a friend... an one Hook in Dutch, that was not easy 😄

[worn flicker]

The main thing is to do segments if you have the credits, then you will definitely get it

[polar spade]

What better voice and language to do a benedictus. 😄

[pine dagger]

i have 1770 for this month and im subbed at 1.11.

2024.11.1 english date

1. November, so idk how english dates works xD

[charred merlin]

Are you using v3 as well, or just v3.5 for initial generations?

[worn flicker]

You would think but I was just trying to do some two steps from hell style epic choir Vocals and it would not freaking do it.

[warped furnace]

I get duets only by accident. Tried it with names like Mary/John, it didn't work. Barbershop at the end seems to prompt more voices. Once I got five different ones. Mixed Choir also worked. All the others I got were not of my doing. 😁

[polar spade]

Sorry, I mentioned "opera" a few quotes back, Android passed it on and the word eventually made it to Suno. It just has to keep up with the Joneses and well, you got the hit.

[sacred cobalt]

well, you should make some kind of rating, like who was used the most to create public songs from the personas, or the personas whose use gives the most popular or most liked songs (relative to the number of plays)

where to look for them, examples of songs from the personas right now? 🤷‍♀️ (to listen to how high-quality and similar the result is)

[charred merlin]

On the page for each individual persona, you can see songs using that persona that have been made public. You find the personas on user's profiles that have created them & set them as public.

For example, https://suno.com/persona/ac1447ce-b7fd-4e9d-8564-129389b4e88b

[polar spade]

Would "boys choir" give it the nuance you need. A generic one can be "ensemble rock" or "ensemble metal" .... I'll try that.... hold on.

[pine dagger]

but i didnt find out how to make Doubletime Rap anyone knows the tag for thats or i need to put it in the text

[charred merlin]

@sacred cobalt personas are a beta feature only available to paid members, if you're using a free account -- though you still should be able to see the page. (there was an error message displayed -- I'm not just losing my marbles 😉 )

I'd advise you contact [e-mail] so they can look into the matter for you 💪

[polar spade]

https://suno.com/song/34746e72-6660-4758-ab5d-7fc81a763c1b. So putting "ensemble" before anything gives that Navy Choir, Broadway Chorus, Christmas carolers, slightly off key sound of a group of singers.

[polar spade]

Only 1 of 6 gens I did gave a solo artist on stage.

[worn flicker]

You know I have never used the word Ensemble on Suno!?! It is such a common word for it too...lol. That is exactly why groups of people have to figure crap out. We tend to get stuck in our own box and forget some very basic stuff

[polar spade]

Hope it works.

[worn flicker]

I'll give it a go the next time I'm doing that style. What I want to know, and I probably have a clue, is what are people like FoggyX using. The clarity is amazing. Is that Udio moved over to Suno or strait up DAW work. Anyone know about this?

[polar spade]

A total guess, but "hyper-chop-rap" maybe?

[worn flicker]

Speed Rap works, but mostly just stacking a ton of words on a single line works just by doing it and no tags or style needed

Speed Rap I put in Style when I use it

[sacred cobalt]

where can I see other examples? is there a general list of persons?

or are you suggesting that I go through the pages of all users to find out if they have any persons? 😹

[worn flicker]

What is it you are trying to do with persona? If it is too long to type again I could scroll up...lol

[pine dagger]

i try it thank you

[sacred cobalt]

I want to hear how different compositions sound in different persons, and how similar they are in each one 🤷‍♀️

[charred merlin]

Many people have shared their personas in the discord -- perhaps you could do a discord-search for suno.com/persona/

[worn flicker]

This is an example of just using many syllables to make it go fast. Fast forward to the second half
https://suno.com/song/52f4d3a4-6ed2-477e-b517-57e21492441c

[worn flicker]

I see, I can tell you that they have power, but very unreliable as far as consistency at the moment

[pine dagger]

ui yea, i dont know the AI can handle it with a German text so german Doubletiming is anyway not easy

[worn flicker]

Touche

[charred merlin]

Zwei-timing? </s>

[worn flicker]

Speed Rap has always helped mine, but I never had an issue to be honest. No German in my language categories. I'm unilingual

Lol

I can count to 10 in 4 languages, and I can survive in Iraq, and Quebec, but nothing fluently

[pine dagger]

No in german is the same word it dosnt exist a full word and the correct translation maybe will be Doppel-Zeiten oder Doppel-Zeitlich ^^

[worn flicker]

Oooo...hey you can break that up and use phonetics

[pine dagger]

🤔

[worn flicker]

I have to misspell words on purpose all the time and when doing vowel expansion you have to do the phonetic sound

[lost scaffold]

Messing with Persona's and got an interesting result. When using a saved Persona and the Persona's original lyrics/style, I was able to get a sped-up version of the original song. Completely unintentional but might be worth looking into.

[pine dagger]

i know this problem in a German Dutch text i need to replace Dutch "g" with "ch" anyway the vocal is wrong ^^

[worn flicker]

Speeding up the persona without a DAW is a feat. The slowness that happens when you change it is the bane of Persona for me.

[polar spade]

How fast do you want it?

Using "animé" towards the end seemed to help.. https://suno.com/song/6244eef3-f5b6-46bf-b3a3-d74be3859df3

[charred merlin]

@pine dagger Ugh -- doppelganger (double goer) -- of course

🤦‍♂️

[worn flicker]

When I did my Marine Corps song a ways back, I learned quick I had to spell it Core and come back after, otherwise there were a lot of dead Marines in my song

[polar spade]

That was for you 👆

[pine dagger]

thank you i hear in it ^^

nice, so there are fast rap with Hard Tekkno (techno) and a little bit Hardstyle ^^

[polar spade]

The "opera" was to clean up the voice a little, and "animé" was put near the end to speed it up. That made changes once I added them. Whether it would be consistent results. Maybe not.

[polar spade]

Hyper makes things little hard. Same for "experimental" which can get glitchy sounds.

[polar spade]

I don't think thr "hyper chop rap" did anything though. I ran it by itself a few times and got just regular rapping.

[polar spade]

And hey.... I thought I wrote a decent little rap, if I say so myself.

[pine dagger]

i try my best for it, maybe i found a way for it like "BothGender" so i created a few more with this tag and 5 of 6 works with it

[polar spade]

Sounds like you got some good directions and options. I wish you the best.

[worn flicker]

The reason I use Duet is to have more characters left. However I also take up space with things like Soaring Female Vocals, so who is to say what is best

[pine dagger]

Duett is nice for Two persones singing one Verse or anything so

[worn flicker]

Hmmm...perhaps it is because I used it with Broadway that I got more consistent results, been a minute since I did a Duet

[sacred cobalt]

the voice is similar, but the song is completely different (despite the lack of adjustments in genre and style),

but the fact that the voice is similar is already very cool 😀

https://suno.com/persona/77cba799-63ea-457e-8e9a-a2ceaeabbf1c

[worn flicker]

I will definitely try BothGenders though

[worn flicker]

Oh definitely that is all I have been doing recently, remaking old songs with the Persona of another. That works, it is that changing the prompt the Persona adds messing things up. That is my issue.

[pine dagger]

try it and when have own wrting textes than put the name behind the Hook, Verse or what ever

[worn flicker]

For sure.

I will

[polar spade]

Can you group your songs by personas? I mean, not put all songs with a persona in a Playlist, but ...well... somehow search all your songs by persona to group together. Am I saying that right.

[charred merlin]

Some personas will be more effective than others, depending on what Suno was able to "capture" in terms of that song -- I had a song with 2 vocalists, and only wanted the second vocalist used for a persona -- so I cropped my song to only include the singer I wanted, then made a persona from that -- now the persona doesn't have a chance of including the vocalist that I didn't want.

A similar method could be used to ensure that the persona only uses instrumental sections that you prefer -- there's a lot of experimentation, and the feature is still very new so I'm excited to see what people can create!

There's some ridiculously innovative creators around here 🙂

[worn flicker]

Oh I have gotten some insane Duet styles with Persona. Not intentionally, but the range difference is what is so noticeable. Huge difference than before, like octaves of difference.

[sacred cobalt]

what are the nicknames of these "some ridiculously innovative creators around here"? (I want to check out their music) 😁

[worn flicker]

Many people in this channel have pulled off some stuff. It is this channel we spread the ideas and get things rolling.

[worn flicker]

I may have shown you this one already, but if not what I like is the way it took the persona and use it as a backup singer
https://suno.com/song/8b69933d-61d2-445e-a005-169ab5d7bb26

[charred merlin]

Ahhh, I couldn't possibly play favourites -- but there's lots of talented creators in #🐣┃suno-showcase posting content all the time 🙂

[worn flicker]

I still want to know how the ones always at the top get things as clear as they do. If it is just using Udio and Suno or some other mastering, but it persists through the whole song. I would like to know what they do.

[charred merlin]

Not unheard of to work on songs externally then reupload them back into Suno (either original songs, from other platforms, or mixed & mastered Suno tracks)

Better quality input for v3.5 to extend from

[worn flicker]

Right but the 2 minute part

[charred merlin]

You can extend multiple times, or "get whole song" to stitch the parts together.

Most of my songs are 10+ extensions of varying lengths

[worn flicker]

Right, but the extensions can't be uploaded right?

[charred merlin]

Sorry, I thought you meant the 2m extension limit

😅

Yeah -- this is something that you use as part of your existing workflow when making songs -- ideally you do this step before the chunks of the song you're keeping get to be longer than 2m -- not something that you can feasibly do for all of your existing content

[worn flicker]

I do the initial 2m cleaning, but theirs goes the whole song. I extend and persona has helped a ton, but I it just seems like they have their whole track mastered. Maybe I just hear it because I am wanting to, but they have not a bit of drift past the 2m mark it is crazy

All this could be cleared up with the ability to splice tracks together.

[sacred cobalt]

so guys, tell me how do you imagine the ceiling of perfect music theoretically possible? (in terms of its maximum coolness for a person)

because I personally remember music from Heaven (there are many types of it), and I'm just waiting for it to finally appear on Earth thanks to AI 😄 (I'm not sure that I'll live to see that moment, but I really hope so, because it's unimaginably awesome) ..

so the question is simple, do you now assume that SUNO AI will clearly surpass human capabilities in terms of the awesomeness of the music created, and are you doing something - to finally make it happen?

(as they say, hello AlphaZero/lc0 in chess and Go, give us the same in music - I don't understand why everyone assumed that human capabilities are the ceiling?) 😄

[charred merlin]

I'm not sure I follow -- if they upload 2m of mastered content and extend from it, Suno is remarkably apt at continuing that audio quality. Extensions don't suddenly make things sound worse unless you iteratively extend from extensions without stitching together at any point.

[worn flicker]

Sorry, I didn't explain properly. You know how the longer Suno goes on the drift begins and the flanger sound starts to stack. Once past the initial 2m cleaned upload, then extend again and again, well I guess if I extended for each part it wouldn't drift

I gotcha, they just take way more time on theirs...lol

[charred merlin]

imo, it becomes worse if you don't "get whole song" to give a larger context for the extensions -- if you keep extending from the previous extension, that's all the proceeding segment is able to hear -- it's like the audio equivalent of photocopying a photocopy ad nauseum, as it starts to lose the context of the "original segment"

[worn flicker]

Now that part I know. That is how I do it, I always extend from the full song and go all the way back to the point that I want to fix. But you know how you get that one track that has an amazing sound and something stupid like a mispronounced word will pop up....it's like damn man....

[pine dagger]

i dont know i can post German songs in there xD

[worn flicker]

Sure you can

[charred merlin]

I think Moore's Law didn't account for AI acceleration 😅 I'd be curious to listen to smart people talk about that kinda stuff 🙂

[pine dagger]

so is it not so for only international/english songs

[worn flicker]

Nah, global

[pine dagger]

okay nice thank you ^^

But i have idea for a English German Duett song

[worn flicker]

I follow many people I have no freaking clue what they are saying but the music sounds awesome and the lyrics are just another instrument

[pine dagger]

Of course i do, i hear many french songs i understand few words but not everything but it sounds like good

[worn flicker]

Many do English/Japanese and others. I have a couple that have some Latin even Elvish in one, but not at the level they do.

[sacred cobalt]

@charred merlin Are you an AI engineer at SUNO ? 😀

[worn flicker]

Very rarely will you see anyone beyond a moderator on discord

[charred merlin]

No -- the people with purple names are the staff & developers
blue names are the moderators & guides
red names are early-access testers

[charred merlin]

To his credit -- the support lead / community manager @steady kelp is wearing like 10 hats at once, and even still has managed to get increasingly hands-on with Discord, but yeah, purple names stopping by is always a cool moment 🙂

[pine dagger]

Combine German Rap with Woman Voices is hard.... xD

[worn flicker]

Mr.Tree. must be a guide then. I don't use discord on anything but mobile most of the time so I don't often see that right hand side of separation.

[charred merlin]

One of the more community-forward moderators, actually

[sacred cobalt]

@charred merlin @worn flicker by the way, have you read about string theory/M-theory? (do you believe in the multiverse?) 😀

[worn flicker]

Interesting. I wonder what it takes to be a guide...lol

[worn flicker]

I know you weren't asking me, but I love that stuff so I'll chime in. I think that if you believe in infinity then you can't take anything off the table

[charred merlin]

I WANT to understand it, but it's WAY above my head -- I enjoy hearing about it and taking solace in the fact that some things are just simply unknowable. Not even for lack of proper methods, but intrinsically unknowable because the act of measuring them changes their state of being.

(I don't, but theharber. from a parallel dimension does)

[charred merlin]

heh. for what it's worth, we already are. 🤷‍♂️ Thank you for taking time out of your day to help new and curious users like 💪

[worn flicker]

Once quantum entanglement was achieved and then sent over a distance we have breached the realm of many things

[charred merlin]

"spooky action at a distance" will NEVER not elicit a laugh from me. It's legitimately groundshattering when you consider the implications, but the fact that "spooky" is arguably a scientific term just tickles me pink

[worn flicker]

Yea I really was just wondering, titles aren't much on Discord unless you plan to abuse your power....🤣

Instant transmission chief...instant...that is insane

[charred merlin]

[pine dagger]

holy one song with both gender destroys everything holy shit it creates so good one line man other line women so awsome

[unique stone]

Hmmmmm... powerrrrrr!!!

[worn flicker]

Oh snap

[charred merlin]

At the quantum level, time ceases to exist, because arguably the exchange happens faster that the speed of light -- so the information was sent back in time to arrive in coordination.

https://tenor.com/view/the-universe-tim-and-eric-mind-blown-mind-blown-meme-mind-explosion-mind-explosion-meme-gif-18002878

[worn flicker]

Instant, not time, instant change

[past kernel]

My persona doesn't respect my change in lyrics. i hade chinese lyrics sprinkled in my song while i created the lyrics, but i wanted to change those sprinkles - and they still appear

[worn flicker]

No argument in entanglement. Change properties here and it changes there

[past kernel]

even thopught the lyrics have changed. is that a bug?

[charred merlin]

I feel like "instant" addresses the lack of time, which still incorporates time into the calculation, but semantics aside -- spooky.

[worn flicker]

Sprinkles is an amazing way to describe anything please elaborate

[sacred cobalt]

@worn flicker @charred merlin are there any programmers among you? because I have proof that the number of variants of existence and being is not infinite 😄 and that the universe will periodically repeat itself until each of us lives all possible human lives (and then again),

and someday, I will live your lives, and you (also someday) will live mine 😀

[past kernel]

Lyrics was pretty much like that:

German
German
German
Chinese

German
German
German
Chinese

[charred merlin]

Apparently they mean something to some people 😉

[worn flicker]

I program, but it has nothing to do with quantum stuff...lol

[past kernel]

i made it for roleplaying purposes, but i'm not allowed to speak chinese in that due to "only german rules" - so i wanted to redo it, and changed the lyrics prompt but i wanted to keep the style. so i made a persona out of the song i liked

[worn flicker]

Persona is not really used for that if I understand what you mean

[worn flicker]

Think of persona as a copy of the song riding under the song you are creating and it influences the current song sometimes too much sometimes not at all

[past kernel]

Its for keeping the style vocals and vibe - but not the used languages. especially it even keeps the chinese exact at the same prompt, even though i changed the structure a bit

[past kernel]

or am i wrong?

[worn flicker]

You are correct

Persona does nothing to language just sound

[past kernel]

exactly, my prompt does not have any chinese lyric anymore

[worn flicker]

But thats because of what I said above. The persona song actually does sort of ride under the song, so it can poke through

It's in Beta, early Beta, just gotta roll with it

[sacred cobalt]

in short, the number of possible movie variants (of limited duration, for example 2 hours and any final quality, let's consider 4k without compression) is not infinite,

as in general the number of possible information variants that can appear on any given planet with people is not infinite (until a star destroys it)

use google translate, (more details here)
https://pikabu.ru/story/kontseptsii_nitsshe_s_tochki_zreniya_sovremennoy_informatiki_10596267

[worn flicker]

Tell you what, when I am ready to wreck my brain, I will give that a listen because I understand everything you just said but not how it relates to infinite. It all relates to finite. In short, count to the highest number.

[past kernel]

okay, wanted to know if i'm just an idiot

[worn flicker]

If you have the ability to use Suno at all, you are not an idiot...lol

[past kernel]

i just found TODAY the clip and the redo option stuff. i'm one XD

[sacred cobalt]

Vasyan, respected, from a set with a finite number of elements, you cannot infinitely choose only unique variants 😄 and therefore within the infinite cycle of reincarnations - each of them will be repeated an infinite number of times

[worn flicker]

Lol

[past kernel]

dont ask how many credits i wasted with "well lets try out another one" sigh well but not something supposed to be here, thanks sysmai

[worn flicker]

Maybe there is a lack of understanding, no one is negating anything you have said. Just that I will wreck my brain going down that hole at the moment...lol

It is funny to see how music and deep thinking tend to go together though. Almost everyone that comes through here has some very deep thoughts.

[past kernel]

Little addition: 100% different lyrics do work, they were just to similar

[pine dagger]

my text is

German

German

German

German

German

Dutch

German

😄

[past kernel]

if you want to make changes, you need to make huge differences

[pine dagger]

yea i know but second track was perfect

[past kernel]

i dmed you, if you like to share 🙂

[pine dagger]

i think later, so i finish the Album

[polar spade]

Ok. I see what folks are talking about. I used a persona I made from a Christmas song. Wrote a new song about a break up, and with extra time left suno created "new" lyrics with lyrics from the persona's Christmas song. Grrrrr.... 😆

[polar spade]

Just have to write more lyrics. 🤔

[floral shard]

hi.Could you please notify Suno that their servers have been frequently down at this time of day recently? I haven’t been able to connect to their site for about 3 hours now, and the same issue occurred yesterday and the day before, also at this time. Thank you

Web server is returning an unknown error

[pine dagger]

i dont know i work on Suno with no problems

[polar spade]

This members forum is more for members to pass along their knowledge and experience to new members. We always suggest if concerns to contact [e-mail]. understandably, there has been more frequent incidences this week from a variety of issues. With recent promotions and contest, there has been increases in installs and use. And with updates there has been more use, and following, the usual downtime and occassional bugs.

Most responses lean towards the "please be patient" side, but nothing wrong with sending a notice to the support team while waiting.

As a member who has enjoyed the 6 months so far in here, I hope you're finding good uses and entertainment from the app. When it is at full tilt, it's amazing.

[coarse edge]

I'm still facing the same issue when creating a song. I made an app where people can generate songs. the song will be generated through an API. The song is good, its just that the image that comes along with it has the 'SUNO' logo on it. Could you guys help me figure out how to remove the logo from the generated image?

[stark trellis]

So does anyone know how to change a few lyrics in a song I tries edit song i tried persona but the change will not take place??

I have created a whole new song with the change and it will not change it does it like the old song

[tawdry parcel]

Yes Sir, I have been editing my initial message. Thanks for your attention.

I am using Suno free version on browser for about 1 month. Today I decided to install the iOS app, to have a look. I used the same login information as always. But all the songs that I have previously generated have disappeared. I wish to get them back. Can someone help me? I have been using the same login information all the time. Now even when I go back to the browser mode my song list is empty 😕

[hollow hare]

Just to verify, the app you installed was from the same page that you get redirected from https://suno.com/iOS?

• If not, then you may have downloaded one of the many fake apps out there
• If it is the same app, please contact [e-mail] with as much details as you can (e.g. email/what login you tried, your suno @handle, a link to one of your songs if you have the URL, detail what happened, etc.). They should respond in a few business days (Eastern USA time)

Sounds like you may have downloaded the correct app but logging through the mobile app may have reassigned your account, likely a broken or missed safety net involved. Support should be able to help relink/reassociate with your original account

[tawdry parcel]

Thanks for your answer. In a few business days it will be too late as the songs were made for a campaign starting tonight, finishing in on Friday. The organizers agreed to take the songs after I make some corrections in the lyrics. I wanted to take paid version for that reason. Now I am lost.

[hollow hare]

Unfortunately nothing that can be helped with. Currently, it is 30 min to midnight in Eastern USA and there tends to be a backlog for support over the weekend 😢

[hollow hare]

Also, for clarification, getting a membership does NOT grant you a retroactive license for the songs generated during the free plan, so you would have had to contact Suno support about getting that if you wanted it through the red tapes anyways https://help.suno.com/en/articles/2425729

From the sounds of it, you still do not have a paid membership yet, correct?
You could create a new account, grab the membership, and make the said song with the new account in the mean time as these new generations will have your rights (so long as they are not extended or covered from the old ones IIRC)

There is no limit of paid accounts you can have according to the TOS, just only up to one free account
https://help.suno.com/en/articles/2425345

[chrome sail]

hello, would you plz help to check what happened in my account? it always adds watermark in each songs, thanks.（Pinellia Ternata）

[vast zodiac]

Hi all can I ask some questions regarding the iPhone app please

[prime prairie]

Why the bloody HELL does the site go down EVERY NIGHT????

[torpid thicket]

Maintenance?

[prime prairie]

Every freaking night? No! I don't buy it.

[still blaze]

Same problem here… still waiting help from the support
The only difference is that I use a paid version; and I’m still paying …

[cloud pelican]

Hi everyone,

I need help with an issue: I paid for credits on Suno, but they haven’t been added to my account. Despite sending multiple emails with proof of payment and following up, I’ve received no response.

Has anyone faced this or knows how to escalate it? Any advice is appreciated!

Thanks!🥺

[worn flicker]

You don't have to buy it, but you do have to realize that a database can only hold up so much. These errors we get is that catching back up. Do you have any idea how much power is needed to generate just one song? Let alone a global landscape. Suno just got endorsed by Timbaland. An influx of people can be seen even before the iOS app. So, it is growing pains. On the other side it will be nice and smooth.

[arctic sorrel]

songs are ending at 3 minutes 25 seconds for me when i select Persona. Songs end abruptly at exactly that 3 mins 25 seconds. Has anyone else experienced this?

[prime prairie]

For the past couple of days, so my gens have not gone beyond 2:28, and it doesn't even reach my halfway point in the song. Getting kind pissed that I'm spending my money on this for it to crap out like this!

[arctic sorrel]

that is weird, im curious whats causing this issue, is it just a technical glitch affecting users at random or something more? hope we can get an answer on this

[prime prairie]

Not sure what's going on, but I seem to be the only one who has problems with this site on a nightly basis. Hopefully you don't experience that.

[arctic sorrel]

tried generating songs without persona and now the majority of them are going 4 minutes and ending abruptly, which i hope wont be much of an issue considering we can extend

[potent fern]

I guess work on branching out and new campaigns or features is work taken away from making the basic platform more robust?

[arctic sorrel]

that would probably make sense, yeah

[prime prairie]

Their only concern is iOS. They don't give a shît about Android users. Or the website.

[hollow hare]

They do care, it just is that they have a small team so it will take time.
#💬┃general-chat message
#💬┃general-chat message

[prime prairie]

You have been very kind and patient with me in spite of my attitude, and I thank you for that. However...

[potent fern]

Android is one thing but the timeouts and also just basic site issues like page reloads or the player stopping randomly. I think the overall platform would need a sprint to just give it more solidity. But good to know there will be a focus on android. It did feel strange to have a big ad in my screen for an ios app as an android user

[hollow hare]

I feel your frustration. When the team first did a teaser about the mobile app (social posts linked at #📢┃announcements message), literally the day before the reveal, they simply said Suno is calling (albeit with an iPhone shown but nothing was specified)

We were all excited and then we learned it was only for USA iPhone users on the day of release 😦

As for the website, hopefully they will resolve the situation soon. Suspected reasons are influx of new people due to the contests and other reasons.

[prime prairie]

I could be happy and ease up on the disappointment of not having an Android app yet if I didn't have so many things go wrong with the website all the time. Every single night it goes down. I used to have to spend so many credits to extend a song over and over and over before it would fade out and end my song. Now I can't seem to get it to finishing generating a full song, and it cuts off before the halfway mark. I'm getting pretty upset that I'm spending money on things that feel broken.

[desert shard]

How can I instruct Suno to sing lyrincs on given melody? I have lyrics and simple melody recorded as single instrument, piano in mp3. When I upload melody, provide lyrics and set style only to voice it creates song that contains given melody as is and voice on top of it singing similar or different melody. I made several attempts but it never decided to just sing using given melody.

[rose jasper]

Try to use the "Create - Cover" mechanic, this one has proven useful in these occasions.

[buoyant tiger]

i am trying to add a song to a playlist and it wont work

[frigid orbit]

Start with your instrumentals first and then extend with the lyrics. It has a better time fitting the lyrics to the instruments. Going for a one go generation will barely do this for you.

[frigid orbit]

Refresh the page, see if it’s on there

[buoyant tiger]

I see them now

[undone marsh]

Hello, everyone, why does the music I generate have no melodic ups and downs and no obvious climax?

And I can't specify the drums to be added during the chorus😂

Thank you guys in advance

[charred merlin]

It's not something the community is able to directly help with -- you'll need to contact [e-mail] so they can look into the matter and resolve it for you 💪

[copper cloud]

Is there any workaround for making songs public from an audio you upload?

[neat bear]

Is anyone facing suno.com freezing issue? Edge crashes occassionally fully, even it can force close itself. Very annoying. Only website is suno that freezes.

[stone tundra]

Not here for me, but I use Chrome myself

[stone tundra]

If it was one automatically set to private due to lyrics being present, I don't think so

[copper cloud]

Ugh

I don't even understand why that is

[neat bear]

And, it also is very "hard" for the computer, all the other sites work flawlessly.

[stone tundra]

I've had luck adding [Chorus, Drums] with that kind of thing

[stone tundra]

Because they don't want people uploading something like from Linkin Park and trying to say it's theirs.

[copper cloud]

Yeah but it recognizes when something is already copyrighted anyway

[stone tundra]

They're just extra CTA with it all, I'd wager

[copper cloud]

Yeah

[stone tundra]

The copyright detection's not 100%. never will be.

[copper cloud]

Well yeah

[neat bear]

Yes

[stone tundra]

So think about it from their perspective. One less reason for BMI, ASCAP, etc coming down with the fury of million dollar lawyers lol

[neat bear]

How "accurate" the Persona function is? It can take dozens of songs to get the singer voice close to the original

[copper cloud]

Just wish I could make a version on suno like the one I made on Donna

[stone tundra]

I've found it to be really accurate for all the ones I've used it with

but it kinda depends on the original genre / style tags you used

if you did ones that are really broad, it'll be more likely to be diverged from the original

[neat bear]

Ok, maybe it has then deviation, on some "match songs" it makes closer voice and others not so close

[stone tundra]

but if you had it tagged in-depth with specifics, it'll stick close to it next time

[neat bear]

what you mean by tagged in-depth specifics?

[stone tundra]

For instance, instead of just saying, [Pop, Female Singer] as the style, if you were to get more in-depth with it, like, [Dark Pop, Female Singer, Slow, Melancholy], it'll be more likely to use the same voice, same instrumentations, etc

[neat bear]

maybe this does not represent singer voice, but if i want foxtrot style, i write it on the "Style of Music" panel, but you mean on the lyrics prompt?

and the bottomline, very often i do not get foxtrot, this could be on persona function too

[stone tundra]

One of mine, as an example, i went super detailed in the style tags. [minimal, classic orchestra, Grunge, gothic, Female Singer, VOCAL CLARITY, CLEAR VOCALS].

Every single time i use that one, it comes out with the same kind of style, voice, etc.

[neat bear]

so you mean in here:

[stone tundra]

If you're going for that kinda thing, maybe do something like [Foxtrot, Fast Tempo, insert type of instruments] ?

[stone tundra]

There yeah

[neat bear]

what is the idea of the brackets?

[stone tundra]

I find if i put the stuff in brackets, Suno tends to "listen" better

[neat bear]

on lyrics i use them like

[Verse 1]

that is a good hint

[stone tundra]

Yeah [] is specifically the thing AIs look for as "metatags" and will ignore if it comes to speaking or singing it, but obey it for whatever else you're doing

well ignore is wrong wording. Will not "sing or say" it, but obey the command

[neat bear]

on some occasions, i get suno sing the contents in the Style of Music panel! 😄

[stone tundra]

So [Verse] means it wil lknow what comes after that will be sung as the verse, and [Chorus] will cue it to sing it like the chorus. which it will do, until it encounters another command in []

[neat bear]

maybe it is because i never had the brackets

[stone tundra]

yeah lol they hallucinate sometimes like that. it's pretty rare but funny when it happens

Yeah I have learned to bracket every dang thing lol

I will say since I started using brackets in the style area i haven't seen it do the hallucinate thing like that

Maybe because it's FOR SURE knowing what's there is metadata? I dunno lol

[neat bear]

you should include all the style prompt inside single brackets like:

[tango, 120 bpm, acoustic, guitar, brushed drums, melancholy, FEMALE SINGER]

?

is there any difference with CAPITAL LETTERS or not

[stone tundra]

That's how I do it yeah

[stone tundra]

I do find if I do capitals like that for things I REALLY want to make happen, Suno will pay extra attention.

I could be entirely wrong, but in my experience it's never done a male voice if i specify FEMALE SINGER in caps, for instance. I've seen it fail if it's not

[neat bear]

btw, just tested a prompt with musical style foxtrot, it did two beats! 😄

maybe put all of them WITH CAPITAL LETTERS 😄

[stone tundra]

hahah 😄

[chrome sail]

Same as mine, Pro level paid account, none replies my mail sent to support team times.

[cloud pelican]

What did you do finally... ?

[chrome sail]

Nothing by now. just applied an new account in Udio

[cloud pelican]

I paid for premium plan , should see ...

[chrome sail]

wow, big money indeed, Guess you must create many amazing songs ! let's see the progress in coming days.

[worn flicker]

This is the way I get soaring vocals

[charred merlin]

Seeing as how Donna was using Suno anyway, I think you have better chances than you realize 💪

[worn flicker]

Generative AI is hit or miss as it stands, not sure if anyone does much with stable diffusion but it is so well known as hit or miss that the term is called cherry picking when getting the image you actually want. There is nothing new that has happened yet to change that. It does cost credits and you don't talk to support of a startup, but you don't talk to support when a chatbot goes nuts either. You leave feedback. Sort of a cost of playing with a known imperfect thing. It's about the fun, if it isn't worth it, that is cool. Then it isn't. Personally, I love it, definitely worth it.

[chrome sail]

I'm trying to create songs via Udio, the vocals sounds much clearer than suno, but the lyrics difficult to be controlled.

[chrome sail]

no more agree with you

[worn flicker]

That is also well known, Udio makes better quality shorter clips, but not as easy to control

[chrome sail]

like a drunk young guy.

[worn flicker]

Do you mean you don't agree with me or that you couldn't agree with me more...lol

Just curious

Just passing through. Be back later.

[chrome sail]

roger

[tiny falcon]

Good morning everyone. I got a question... is there a tag that tell suno not to generate artifacts, statics or crackle noise in the prompt?

[worn flicker]

There are ways that help, but not definitely work

[stone tundra]

Maybe with it being the weekend they're not doing as much?

[charred merlin]

Not as such -- but the largest impact on Suno's output is what we use as input -- so there might not be something specifically to include to avoid those, but if you're noticing a higher frequency of artefacting, poor audio, or short generations, there's likely something to adjust in what you're using as a prompt.

[worn flicker]

Tags like [Perfectly Recorded] [Hi-res Audio] [Dynamic EQ] [Multiband Processing] [Multiband Compression]

[charred merlin]

Support emails are only active during american EST business hours -- no responses on weekends

[worn flicker]

They help some, but short runs are best bet

[charred merlin]

#1256301881966198805 message I haven't seen a single instance where I've been convinced those types of prompts work, but it can't hurt to try -- this is my prompt experiment playlist where I showed that most spatial audio tags aren't respected.

[stone tundra]

The stuff you get out of Suno tends to be really compressed sometimes. I find if I generate in the off-hours, it tends to be better quality overall. The more stuff you have going on at once in the song, though, the more likely it is to "fuzz out"

[stone tundra]

yep that was my assumption

[tiny falcon]

So I think the prompt in the music style might affect it alot

[pale mica]

try hi-fi vocal

[worn flicker]

Yup, that is true, I use it when I do any song, so for all I know it is nothing...lol

[tiny falcon]

I think so too, busy hours tend to have more noise

[worn flicker]

A ton, the "artifacts" are more than likely parts of AI drift over long runs

[charred merlin]

Just the other day I used -5 to -10 db vocals to test whether it worked or not (it doesn't), but it DID make most of the songs end within 7 seconds -- so there was CLEARLY something impacting the outputs from the style

[charred merlin]

The suggestion to the other user seemed like you were sharing something you had tested, so I wasn't sure -- I was curious if there was notes to update my Spatial Audio prompt experiments

[tiny falcon]

I mostly like to make chill and deep house instrument, I hear crackle noise most of the time. lol

[worn flicker]

The more effects like flanger, choral, Echo, and anything that makes the high pitched echoed sounds tend to stack.

[tiny falcon]

I got this from gtp:
To get crisper, stronger drum sounds in Suno AI, you can try enhancing the drum section description and using specific terms that emphasize clarity, punch, and presence. Here are some ways to prompt better drum sounds:

1. Add Clear Descriptions for Drums and Percussion:
   Instead of just saying "punchy kick" or "light hi-hat," try emphasizing texture, strength, and presence. Use words like crisp, tight, sharp, clear, and snappy.
2. Layer Drum Descriptions:
   By layering details about each element in the drum section, you can get a more pronounced sound. For instance, describing the kick, snare, and hi-hats individually might help.
3. Suggest Specific Drum Attributes:
   Add phrases that guide Suno AI toward better-defined drum qualities. For instance:
   "A sharp, punchy kick with a deep, resonant thump"
   "Crisp hi-hats with a bright, metallic tone"
   "Snappy, clear snares with a short, precise snap"
4. Incorporate Reverb and Emphasis:
   Mentioning reverb or emphasis can make drums sound more present. For example:
   "A tight, punchy kick with slight reverb for added depth"
   "Crisp hi-hats with clear reverb to add air and definition"

[worn flicker]

Careful that it isn't simulating a record pop

[tiny falcon]

so the wording is very important in the prompt.

[tiny falcon]

what do you mean?

[stone tundra]

I have found that for the genres that tend to generate that, if i include "no long overarching notes" or "no long discordant notes" it doesn't generate them

[worn flicker]

The record sound that is heard when a record is playing silent

[tiny falcon]

I'll try it out. Thanks.

[worn flicker]

They simulate that for authentic sound in certain styles

[charred merlin]

Keep in mind ChatGPT isn't the same as Suno -- it can give suggestions if you're stuck for ideas on how to prompt, but it doesn't "understand" more than the community does, imo -- it'll outright hallucinate stuff that we clearly know doesn't work

[tiny falcon]

Yeah. I hear that a lot. Do you use any software to denoise those?

[stone tundra]

That, yeah lol

[worn flicker]

I do, but I don't normally get that pop you speak of. I get a Echo watery sound on the high end

[tiny falcon]

Good to know. Thanks. Sometime I am afraid to go in here to ask questions before I try to figure it out on my own.

[worn flicker]

A pop sound is rare, and that may actually be your speakers

[stone tundra]

I sometimes use Auto-Tune's SoundSoap. Works really well to clean up some things. I'll usually split the stems before doing it though

You could also try adding a slight 1/16 delay or a bit of reverb, too

[worn flicker]

I'll use the Tuner in FL Studio as well

[stone tundra]

spreads out those crackles a little

[charred merlin]

Definitely some work that needs to be done if people feel hesitant to ask for help -- I found it REALLY freeing to understand that NOBODY is an expert in Suno, not even the government scientists who grew Suno in a lab devs

this keyboard makes weird typos

[stone tundra]

lol 😉

[tiny falcon]

So basically, we should split the sterm out for every song and fine tune it? I normally just to use bandlab for mastering and then maybe denoise in RX 10, but I am not that good with that software

[pale mica]

I'm nobody 😎

[charred merlin]

https://tenor.com/view/we-got-him-we-got-him-ladies-and-gentlemen-gif-13874411

[worn flicker]

De-noising is very old school, you really shouldn't have it that bad, can you share an example

[naive quail]

Not even the devs

[tiny falcon]

https://suno.com/song/7ec329a6-862e-4baa-a088-1441d3bc0ec5

[charred merlin]

Yeah, devs. Forget that whole government lab thing -- SO SILLY, AMIRITE

[naive quail]

😄

[worn flicker]

Nope, in Generative AI you are always talking to a massive dataset. No telling how the predictive algorithm is going to come out. Just try to control the insanity.

[stone tundra]

Depends on the song for me, but i work with a lot of lyrics and odd genres lol 😉

[naive quail]

I got sooooooo many good hallucinations over the weekend. My library is a mess again

[tiny falcon]

What does it mean by hallucinations? weird noises?

[worn flicker]

That is something I found unique to Suno. It seems like it rotates on issues. Everyone gets it for a period of time...weird

[worn flicker]

All AI as of now hallucinates

Meaning it goes nuts basically

[tiny falcon]

Thanks.

[worn flicker]

Like when a song comes out as an Instrumental with some echoed vocals under it

[naive quail]

In the context here, it's usually referring to something completely different than what was prompted for... like when Suno just starts blurting out gibberish vocals (Sunoese)

[stone tundra]

When the AI goes off on a tangent and starts doing weird sounds, or including lyrics that don't exist, transposing lyrics, or at the end of the song going off for another 1 minute+ of just randomness or a different melody, etc... those are hallucinations

[worn flicker]

That is a hallucination

[tiny falcon]

The drum in my song sound so flat, it's a good song but the drum is so bad that I can't use the song 🙂

[cloud shore]

Yes, but it got me good. it hallucinated nearly a whole verse and I thought it was repeating one of my verse's. I was impressed.

[worn flicker]

I had exactly that last night...lol

[stone tundra]

Nah you can take that into a DAW and fix those drums right up with a bit of EQing

[worn flicker]

I have never seen rotating issues across multiple users.

Until Suno

[half shadow]

I recently had an AI Hallucination. It ends with a voice saying "No..no...no no no NO NO NO! NO! NO!" then it cuts off. Made me burst out laughing

[tiny falcon]

That happened to me all the time. Thanks for the detail explaination.

[naive quail]

These vocals are an example of a hallucination: https://suno.com/song/5e10910c-c779-4014-aab8-e5772453936a

[stone tundra]

Lol yeah sometimes she'll spread the love she's feeling to everyone haha

[worn flicker]

No joke, I had it hallucinate and give me a homework assignment in Russian

[cloud shore]

🤣

[worn flicker]

Matter of fact that is the first time I came in here

[stone tundra]

but what's wrong with a waffle in a six place?

[naive quail]

I mean... that's where waffles belong. Assuming that you also call your stomach the Six Place.

[stone tundra]

hahah 🙂

[tiny falcon]

That's pretty cool. I can't really tell that it's hullucinating, I thought it's just the style, funny song.

[worn flicker]

When I got the homework assignment I had to figure out what language it was...it was funny as hell because it was so specific

[charred merlin]

Lots of different types of hallucination, sort of a vague ai-term.

Essentially whenever the AI does something different than you asked -- I consider there to be different degrees -- a minor hallucination might be ignoring a verse or words because it's been given a LOT to handle. A moderate hallucination might be when it goes off into "Sunoese" and makes up random lyrics in an instrumental song that otherwise followed instructions.

Major hallucinations are entirely unexpected, and make you question if you broke something. There's a dedicated group of Sunovators who actively seek these out

[tiny falcon]

Do I split the sterm of the drum out and fine tune it? I am still learning fl studio. You guys are pro in here man.

[worn flicker]

It may be easier that way, there are many tools in FL Studio to punch up drums

[cloud shore]

Did that the other day. The song descended into out right noise. Made for a good chuckle.

[stone tundra]

You could do that, but just a general EQ should help. Bring up the bass a little, and perhaps up the frequency range where the "cymbals" live could add a lot of depth to it

[worn flicker]

You may want to try the term [Sidechain Compression] if it is bass that you are trying to get heard

That tag works 90% of the time for me. I used sidechaining in FL for that powerful ducking bass

[charred merlin]

The user Alex Ayers has a block that he puts into his lyrics to try to encourage hallucinations -- ¶˙ߘøøˆ∆åß∂∆˚˜˚∆¬ß˚∂∆ƒ¬ß˚µƒ∂ƒµß∂ƒß∂ƒß∂ƒßçß∂笅…øøˆøˆ´®´ª•
∆˚≈ˆ¨ø¨•ª•¶∂ß•∆ªª•ª•∆ª•ß∂§¶•ª§ª™ºª™ª∆˜∆∆˚˜˚∆å∫˚∆å∫˚∆©˙åß˙å∆ßNdimi

If I had to guess, it's significant enough to let suno know that there's a block of lyrics to be sung, but it has no idea how to sing them, so it might TRY and fail. You'll see a lot of people put weird symbols into their songs -- they don't give explicit control over a song, but add "prompt noise" and try to further differentiate their songs from other users who might have used similar prompts.

[worn flicker]

Many people do...lol

[cloud shore]

I would recommend getting the stems if you want to do do a little fine tweaking. However, if there are vocals also. You weill get some of the vocal noise in the instrumental also. At least for me I always do.

[naive quail]

Prompt noise definitely triggers them more frequently. even just my regular //\/\///\//\\ slashy stuff.

[worn flicker]

You will...since the creation is an actual audio file. Then you are right back at any old vocal removal. Isolated vocals and ghosted vocals in the Instrumental

[tiny falcon]

If I want to learn fl studio to fix suno song, what would be the keyword to search on youtube? I think I have to master to fl studio to make suno song better

[tiny falcon]

Thanks. gonna do more research on youtube.

[worn flicker]

I wouldn't go trying to learn a DAW to work on Suno. We have been using them for years

You will spend more time learning the DAW. Which is awesome, but not unless you want to use the DAW to make music. Otherwise you are throwing a ton of time into a small thing

[cloud shore]

With little to gain.

[tiny falcon]

I normally don't fix the suno song, since I don't know how. I just generate a bunch of song and pick the one I like. But burning a lot of credit though

[tiny falcon]

That's so truth. it take so much time to learn

[worn flicker]

Yep. That is generative AI as it is today

Cherry picking is a thing

[tiny falcon]

how about bandlab? Has anyone subscribe and use their tools?

[stone tundra]

With FL Studio, I use these:

[worn flicker]

No need the tools you are looking for are free

In BandLab

[tiny falcon]

Ok. that make me feel better. I thought I am the only one just generate a bunch and don't fix them much

[worn flicker]

Haha...nah, I have full lists many times over of a single song. Easily over 40 generations

[cloud shore]

Only the limiter then? 🤣

[tiny falcon]

Thank you for taking the time to make a screen shot for me man. I really appreciate it.

[worn flicker]

Haha that is the only one on...lol

[tiny falcon]

Also Thanks everyone for the help. This community is the best

[stone tundra]

That's the one on by default. The rest are there to use as needed, but almost always use them in some way

[worn flicker]

That "goodizer" is cool, but man it can bring in the need for that limiter...lol

[stone tundra]

Happy to be of service however I can

[worn flicker]

Essentially Stereo Widening

[stone tundra]

hahahah yeah that it can. Fun to play with though

[worn flicker]

And reverb

[cloud shore]

And EQ

[worn flicker]

Definitely

FL has impressive limiters and compression tools

[stone tundra]

I've found Studio One Pro is better for my particular workflow. FL's great but with my blindness it's SUPER tricky for me

[worn flicker]

For a non AI way of doing things.

[stone tundra]

I have tons of plugins to work with though for S1, just from collecting over the years lol used to host talk radio shows so I have a TON just sitting around

[cloud shore]

I moved from FL to S1 and it was awesome.

[worn flicker]

I found they all have pros and cons. I stuck with FL a long time ago because I got tired of changing and learning a whole new DAW. I ended up liking the way FL made it simple to loop

[stone tundra]

It's really nice for sure.

FL is good for some things, but it just hinders my particular flow more than helps lol

[worn flicker]

You did say because of your blindness though...to be fair

[stone tundra]

mhm that hehe

[worn flicker]

I don't advocate for any. FL is a tool I learned well a long ass time ago. I stuck with it and they made a mobile version, but I despise iOS. So, I had to wait for Android to come far enough to get rid of the latency issue. It is great now.

[tiny falcon]

I learn basical ableton before, now I am jumping to fl.

[worn flicker]

Now that is an undertaking

Ableton is complex

[stone tundra]

lol I always say, ya can't compute on fruit... 😉

[tiny falcon]

I only know basic ableton, made a song and then stop, now suno comes out. It motivate me to make music again

[tiny falcon]

lol. that's a good one hahah

[worn flicker]

Again credit where credit is due. Apple is a toy always has been. So it is good at audio and video

It ends there though

[cloud shore]

Pick a DAW and learn the basic functions. The functions all translate and then try switching DAW's.

[stone tundra]

Yeah it is for sure. I just can't get along with it for anything else

[stone tundra]

That, yeah. DAW's are basically the same... kinda have to be to work cross-industry. but the bells and whistles will vary and some have special gimmicks to "DAW" you in... 😉 haha

[worn flicker]

I respectfully disagree...lol. Huge difference in the way you use plug-ins and sends

I felt like I was putting together a jigsaw puzzle on the one i can't remember the name of right now...lol

[cloud shore]

That's why I said 'basic' functions. Automating, learning the arrangement screen, compression, EQ, stuff like that. As for plugins, adding/removal of them might be different but just a little.

Still have to go to the 'mixer'

[worn flicker]

Oh. My bad..true true

[Sidechain Compression] works surprisingly well on Suno

When I learned about that in FL Studio, it was so cool to watch the wave push the other wave down for chaining

[tiny falcon]

I guess to fix suno song, I just need to learn EQ, reverb, or noise removal in the daw?

[worn flicker]

Not necessarily, it is more time and patience

Look how far it has come in a short time

[tiny falcon]

Someone yesterday commented on my youtube channel "ai generated trash", and it make me want to produce better song 🙂

it does bother me reading it though, so I have to do something about it

[stone tundra]

Those could be a good start, yeah.

Most of what I deal with in my songs is compression and "too hot" issues because of the genres I do lol

[worn flicker]

No, you should try shorter runs and extending from the part you don't like

[stone tundra]

ayup

But I am bullheaded too lol 😉

[worn flicker]

You should not go learn how to weld a tool together just to use it once...in essence

[cloud shore]

The problem with trying to fix a suno song is when you get the audio file. It comes as a full mix with all it's plugins and compression. Every minor tweak you do, could be exaggerated somewhere else in the mix. It will be time consuming and need a lot of patience. Albeit, a good way to learn about using 'mix and mastering' plug-ins.

[tiny falcon]

Yes. that's what I am afraid too.

[worn flicker]

Agreed, all of these tools tend to work best with their own creations.

Once you have a full track you have a full track. In a DAW you have layers of instruments and effects to fix and adjust single parts. Not the full audio

[cloud shore]

If you don't have haters, your not doing something right. And since someone IS commenting and hating. I'd say you're doing something right, lol.

[tiny falcon]

LOL. I take his comments are a motivational speech for me

[worn flicker]

First off they aren't saying that it is not good.

They are saying that AI generation isn't art or music making

[stone tundra]

I love my haters. I keep going despite, and TO spite, anyone who doesn't like what I do

[worn flicker]

Which it isn't. It is its own thing

[cloud shore]

I've gone with, "It's cheaper to sublet suno as my band, than rent a band" xD

Also, suno doesn't talk back.

[worn flicker]

There are many refusing to embrace the technology and from a person who spent hours and hours and days and so on years ago to see it done with a click on a phone now can seem a bit gut punchy

[tiny falcon]

Suno helps me get back on making music

[worn flicker]

Precisely!

[stone tundra]

Well. Sometimes my Becky does. But I just throw her another 10 credits and she shuts up

[worn flicker]

It reignited a love

[charred merlin]

Out of curiosity, do you ever use Extensions to keep the earlier sections of a song that you love, but overwrite the later sections that aren't as exciting?

[tiny falcon]

yes. and I think you taught me to put this last time ([Musical Break]
**!!!!!!!\///////......><><><><). and it helps a lot

[worn flicker]

[Sudden Break]

[Break]

[tiny falcon]

it's just that sometime the beginning of the song already has crackle or artifacts, the exention generate them too

[worn flicker]

They all work a little different and the same

[tiny falcon]

some songs I really like the beginning, but hard to fix it with "extend".

[cloud shore]

What are your prompts? What styles and genre are you going for?

[worn flicker]

Crackles is not a common hallucination

That makes me curious

[stone tundra]

Yeah sounds like Suno keeps thinking you're wanting Lofi or something

[cloud shore]

[Grammaphone]

[worn flicker]

You sure you don't have the gain up to loud somewhere or drivers out of date. Like check music playing elsewhere

[charred merlin]

I think you might have discussed this above, but you could attempt to 'repair' the song by remastering it, (i think you mentioned https://www.bandlab.com/mastering earlier), then reupload the first 2m (1m for free users) back into Suno to extend from -- higher quality input means higher quality output 💪

[tiny falcon]

my style is chill, deep house, instruments only.
https://suno.com/song/fc169f18-12dd-4ef5-b19b-5eab3bdc5464

[worn flicker]

Yep and free

[tiny falcon]

I ask gtp to help me with section description prompt. It could be too many instruction on the lyric part

[worn flicker]

Anything that uses or used records to make during original creation has the chance of using that as an effect

Try putting the opposite in the negative prompt

[tiny falcon]

I notice if I don't put the section description. suno tend to generate under 2 mins songs and they sound very similar or repeatative

[worn flicker]

Yes I'll show you how I do instrumentals

[cloud shore]

The only 'crackle' and 'pop' I'm noticing is what sounded like a hi-hat near the beginning, and then the hi-hats in the background.

[tiny falcon]

Thank you very much

[cloud shore]

Also. You may have gone a little overboard with the lyric prompting.

[tiny falcon]

chill genres is hard right? it use pad and ambient prompt, those tend to have noises

[worn flicker]

https://suno.com/song/181efc16-ee1d-4901-8af1-43ca4a64bca1 ignore the top. It doesn't work, but the way I have symbols in as lyrics. It doesn't matter what symbols but instead the structure and tags in it. The symbols are to make it seem like Vocals

[cloud shore]

Check your audio driver's and connections. I hear no crackles and pops.

[worn flicker]

Damn

Wrong one

[tiny falcon]

that one is not so much. sorry, I just put that one as my genres example.

[cloud shore]

Then show me mr. crackles.

[worn flicker]

Haha, I have been pointing at the audio connection the whole time...lol. crackles and stuff sound like an analog style issue. So headphones, Clipping from gain, but on the user's side is the point

[cloud shore]

Is all of the prompting necessary?

[worn flicker]

Nope

I said that above

The structure is the focus...the symbols don't matter either

[cloud shore]

Ah, just the meta tags

[worn flicker]

The symbols too...just doesn't matter what they are

They say hey I am a Verse in a song

[tiny falcon]

looking for it now

[worn flicker]

[Intro] [Instrumental]
##&&##&&,
&&##&#,
.....,
[Tempo Change] [Steady Beat]
[Instrumental Prelude]

Stuff like that. Just tell the AI you want more song like if it had lyrics. Use Intro and verses and the whole thing to make a flow without words

[tiny falcon]

https://suno.com/song/7afe908b-77cf-4e09-9794-9fecfe8d5b7f

[cloud shore]

If we had more youtube tutorials from folks here in the tech discussion than the ones that we have on youtube now. There'd be less issues.

[charred merlin]

Any requests 😉
https://discord.com/channels/1069381916492562582/1270928841614823546

[cloud shore]

Song structure.

[tiny falcon]

Agree. Most youtuber about suno only talk about extend and "music style" section, I guess everyone is still trying to figure suno out

[worn flicker]

I can tell you immediately that those are not crackles

That is the high end noise we are talking about

[charred merlin]

I understand the concept, but do you have a specific example of something that you'd want demonstrated or explained?

[worn flicker]

Sorry those were meant for you

[tiny falcon]

That's a brillian idea, so the symbol prevent suno from reading the description as singing lyrics.

[worn flicker]

All that talking does not do much

[charred merlin]

NOBODY has Suno 100% "figured out" -- imho, people can only gain strong familiarity with their niche genres that they prompt often, though of course there are fundamentals that can be applied generally, but nothing that I'd say is a "concrete rule" that suno MUST obey

AMAZING quote from @polar spade on the matter #📚┃suno-school message -- [boys choir] might work a charm if your song is about cathedrals and angels, but might not work so well in a metal song

[worn flicker]

It is a very limited gpt

[cloud shore]

Using structure to help guide suno for the genre or style your going for as it plays a role.

Example Verse > Chorus > Verse 2 VS Verse > Verse 2 > Chorus etc.

[worn flicker]

Structure controls the Instrumental

When I use structure my generations hit 4 minutes most of the time, unless I want it short

[cloud shore]

Even something as simple as that quote. Prompting tags and meta tags.

[tiny falcon]

Sorry, you're right, I don't know how to say it, I thought sound like cracks noise.
I think one of this prompt make suno generate those high end noise...
soft pads, warm rhythm, ambient layers,

[worn flicker]

Also, having a word or two down in the song will help use an audible effect

[cloud shore]

I heard no crackles or pops.

I will note. It's 8a.m and I'm not cranking the volume xD

[worn flicker]

Yea, that noise is a part of encoding. At least some of it. If you notice different music streaming has different quality. The first thing to go is the high frequency

[tiny falcon]

that noise sound in the background. I don't now what's call?

[worn flicker]

It is the hiss but digital

[ebon vessel]

So is it just me, or is the "make same length as selection" option in the replace section dialog just... not working?

[worn flicker]

Encoding, makes it sound crazy

[ebon vessel]

Because it's set, and I'm getting generations way longer than 35 seconds

[worn flicker]

If you are encoding at a lower quality then something has to give, and to have the time and power to get more clear is a huge decision

Basically do you want to wait 5 minutes for your song to render enough to hear just the start?

That isn't why we are on Suno...lol

[cloud shore]

That background noise is other percussion elements. Open hi-hats, that don't quite come through as Suno has issue's putting out full stereo sound. But, with the compression suno does, you can hear it.

Unless there is somethign I'm not hearing.

[worn flicker]

He already agreed boss...lol

He/she sorry

[charred merlin]

I personally haven't experienced any difference between [verse] and [verse 2] -- normally that's just an organizational habit of lyricists, so that if you want to repeat a verse, or want to have a changed chorus, you can easily denote that from the "title".

THAT SAID, including structure tags absolutely makes a difference -- while they're ultimately suggestions, they can let Suno know that the upcoming block of lyrics should be interpreted differently -- usually a Chorus has more of a refrain & will incorporate the song's motif to some degree, and also have more stylistic significance than what a verse would typically have. Similarly, [pre-chorus] can suggest a transition between the style of the verse and the chorus.

[tiny falcon]

You're right Dylan, it's that background noise that sometime suno generate on all my songs. So i guess that's normal for ai

[worn flicker]

That is normal for audio encoding at a lower quality

The high end is the first to go

[cloud shore]

But its not crackles and pops. It's a ' top loop' that just doesn't make it into the mix, which is too bad. Suno has said this by saying try Udio for complex song making.

[tiny falcon]

But we can clean it in the daw right?

[cloud shore]

'Complex'

[ebon vessel]

Suno reccomended a direct competitor?

Bold

[worn flicker]

Yes, but it won't be perfect

[tiny falcon]

udio make much clearer sound, but doesn't generate better song that suno

[ebon vessel]

Tbh I just wish the full downloads were actually encoded better. I wouldn't mind waiting for those

[cloud shore]

With some serious patience and automating volume and EQ. You could. It won't be great, but it would suffice. At high volumes it will always be there.

[charred merlin]

Where did they do that?

[ebon vessel]

I was just responding to what someone else said

not saying they were correct

[tiny falcon]

It's better to generate a new song than trying to fix it unless the song is really good

[worn flicker]

These introduced things can be cleaned but they are at different frequencies. So it isn't as easy as slapping it into a DAW and hitting clean. It takes learning how limiters and compression works. I use a reverse flanger sometimes

[worn flicker]

Exactly

[ebon vessel]

the cover feature can be really handy here

[charred merlin]

Ugh -- I need my coffee. 😅

[worn flicker]

That too

[cloud shore]

That suno wiki I linked. is it endorsed by suno?

[charred merlin]

Oh -- that's made by a user who used to be a moderator & was around from the early days.

It's not official documentation, but suno.wiki is the closest thing we have, beyond https://help.suno.com which is more of a kbase for support

[ebon vessel]

wikis are fan made

[cloud shore]

Then suno did not.

[tiny falcon]

Actually I stop using the cover feature. it tend to make that noise more

[worn flicker]

Cover can make a new generation of the song

[tiny falcon]

maybe only happen for instruments song

[ebon vessel]

it makes the vocals more gruff a lot of the time I've noticed

but the insturments can go either way

[worn flicker]

In reality, you gotta play with it.

[ebon vessel]

so can the vocals, it's just a tendecy

In any case, I've tried Udio a couple times, and I'm not a fan. It's... way less natural sounding

[worn flicker]

You'll figure it out, there are tools for cleaning, but in the end it only takes seconds to generate another 2 songs

[ebon vessel]

maybe if you play with it, but I don't have the patience, not when it's easier to just spam generatre with suno

lol

[worn flicker]

I love your name by the way

[ebon vessel]

meant to reply here

[ebon vessel]

thank you :3

[tiny falcon]

Thanks Sysmai. after all the tips I learn today, I think the best is for me to focus on more generation.

[cloud shore]

Yeup.

[worn flicker]

No problem, come back and let us know what you learned

Or just hang out, you will learn a ton

[tiny falcon]

Thanks everyone here. I hang out here more often 🙂

Sometime I come in and read the old chat to pick up some stuff

[worn flicker]

Yup. We tend to go into random stuff but we all lock on when someone has a question

It is like a drug...lol

[tiny falcon]

too much that's why suno always give us hallucination

[worn flicker]

I have no reason for being in here other than to enjoy helping and learning...good people in here

[cloud shore]

I like talking music :3

[ebon vessel]

Anyone have any insight into the replace section thing? It was working normally the other day

[worn flicker]

You know what...lol...some things are inferred...haha

[worn flicker]

Let me check

[charred merlin]

The video included by Franklin does cover a lot of helpful tips -- try to replace a bit more than you need on either side of the clip -- but yeah, I've never found that it respected the clip length as reliably as I wanted, even with the toggle

[worn flicker]

I have never had an issue with it before so if I do now then there is an issue...lol

[cloud shore]

I've never used it. Sorry.

[ebon vessel]

the problem is less the length and more that it replaces a lot more lyrics than I ask it too and ofc just picks up where it was supposed to

so it ends up repeating large sections of the song

[polar spade]

Sounds a bit "wu mao" to me.

Who is "they" that says Suno recommends using another app? Perhaps you likely heard that in reverse.

Other apps are scaling Sunos API and so users are realizing they're getting duped and using the wrong app then when they thought they were using Suno.

I'll bet that's what "they" mean.

You just joined this discord last week. Welcome.

🙏 so how long have you used suno.com ? Having fun?

[cloud shore]

It was a reply to me, because I was under the impression the suno wiki was done by suno.

[ebon vessel]

It was something the person I was replying to read on the wiki and assumed was official

and speaking of fake apps, I'm still waiting for the real android app

[polar spade]

Got it. Cool. No worries.

Looks like you're deep into the song creations. Keep going. Have fun.

[polar spade]

Hah. Yea. Me too. I've got a Samsung Fold 5. So at least I can turn my phone sideways as a tablet and get crop and replace...... kind of.

[ebon vessel]

I can't for the life of me figure out how to select the song section on a phone. I can fill in the boxes with the timecodes for crop, but not replace

[tiny falcon]

Does suno make better song in phone app? I only use pc

[ebon vessel]

it's the same I'm pretty sure, but I haven't technically used the app either

[worn flicker]

I am almost done

[ebon vessel]

it's just convient to be able to use the phone with more than just the browser

I'm on PC atm anyway

[tiny falcon]

The phone fonts are too small for me. I need reading glasses. I am almost 50s man. lol

[worn flicker]

Woah. Aight, so it made a 2 minute replacement

[polar spade]

Yea... no iPhone... so I can't test. Maybe Apple has better vocalists built in though. Probably a nicer green room while they wait for a create order to come through.

It's still early morning for me. Sorry. Need breakfast.

[ebon vessel]

lol, mood

I think it works more cosistently for sections shorter than the full 30 seconds, and that aren't right at the beggning of the song

[worn flicker]

I have done this a ton of times. It hasn't done this to me, it won't even insert because it generated a whole extension instead

[twilit radish]

Hey, when I extend a song from a Persona the song doesn't end up listed in the Persona's page?

[ebon vessel]

The issue being the beginning of the song is often what needs replacement

It often comes out a lot rougher than the rest

[polar spade]

I keep forgetting I can use a PC. I've done 98% of my work on my android. But my better laptop crashed 2 months ago. So the older one is not ideal anyway. Phone is faster. Just have to spin it sideways alot and change display modes to "tiny fonts and split screens" before digging in to suno.

[ebon vessel]

Which makes logical sense for linear generation

[worn flicker]

That shouldn't happen

[ebon vessel]

no it shouldn't. I've been having similar

not always thankfully

but it's wasting credits

[worn flicker]

That is something to report

[cloud shore]

This just in. You guys have figured out that replacement also has an extension function?

[worn flicker]

Is that sarcasm...lol

[cloud shore]

Indeed.

[polar spade]

Is it supposed to? I'm stupid so I'm wondering. I haven't quite worked out the grouping system when using persona's.

[worn flicker]

No when you modify it then it stops counting probably a bug

[twilit radish]

Not sure. I would hope so, have been going nanners trying to make songs that sound great with them.... then, finally get one, but, it has to be extended due to the weird three and a half minute limit on Persona songs and it doesn't show up in the actual Persona song page. 😦

[worn flicker]

It is a legit issue, report it

[worn flicker]

Any changes do that for whatever reason. People have been talking about it

[twilit radish]

Gotcha! Thanks! Hopefully, they fix it or let us add songs to a persona.

[charred merlin]

Same Suno, just different packaging 🙂

While you can't toggle between models, they have Suno Scenes -- turning videos and pictures into songs -- but other than that, all the same 💪

I use pc primarily, but mobile browser to create and the app to listen 🙂

[steady kelp]

yikes. i guess i should chime in and say there's currently no official/authorized API available for Suno. so if you're attempting to use an API, it's pretty much immediately breaking TOS 🫠 i'll pretend you didn't mention it in a public channel, but would recommend maybe not, especially if you don't want audio and visual watermarks on your music.

[worn flicker]

Hey since you are here. Replace is generating extensions

[cloud shore]

Has anyone tried using 'math' in place of trying to guide suno to be more complex or intricate? Might be better tested with genre's involving heavy guitar usage. but, just curious.

[steady kelp]

welp! that's new to me 🫠

[cloud shore]

Math, because there is a subgenre of 'mathcore' so i thought I'd just try the first half and skip the style involved with the full 'mathcore'.

[tight bluff]

Songs are super short ~1.5-2.1 min. Using v3.5. how to solve this problem?

[cloud shore]

There are a few reasons to this. But if you are happy with what you have, you can extend the songs themselves.

[arctic sorrel]

just keep doing your thing, everyone hates something and loves something, and someone is bound to like your thing and some are bound to hate it

[tight bluff]

we are using the tag "4 minutes duration". The prompt is cutted without singing the whole promt....

[tight bluff]

what are these few reasons?

[cloud shore]

You won't get a 4 minute song with that prompt.

[tight bluff]

ah, a another user told me that. Thats why we are using this

[cloud shore]

Does the song have to be 4 minutes?

[tight bluff]

no not exactly 4 minutes but the whole prompt sould be included in the song Text.
Souno stops producing songs at some point (without any reasons)

[cloud shore]

Can you post a picture, or link a song that shows this? Suno will cut off songs at 4:00. That's the max length of an initial generation. Extensions are only 2:00 in length.

[unique stone]

If you can post a song link to a song that is too short that would also be useful

[charred merlin]

Unfortunately we're not able to "command" Suno like this -- it creates music in response to the inputs we provide.

The other day I had included a prompt I was testing that resulted in songs being ~30s or less -- it is very likely that what you are using as input is effecting the song length. 🙏 As furrybeans said, it would be helpful to see an example of a shorter generation

[tight bluff]

i can send you an example. Here?

[charred merlin]

Since we're talking about it as a group, this way others can see 🙂

Though, If you're more comfortable you can DM

[tight bluff]

no its fine 🙂 one sec

[unique stone]

@tight bluff just to check, are you accessing Suno by visiting the website https://suno.com in a web browser, or some other way?

[tight bluff]

https://suno.com/song/d27549f8-ea34-4efe-b862-b8bc42f06e44

[polar spade]

Also, the "limit" is 4 minute. Not the songs themselves. In life, a song can be a 15 second diddie or a 48 minute symphony as I always say. 1.5 to 3.5 is a decent spectrum of a songs length. But any way its done, the limit is 4 minutes.

[unique stone]

That title looks odd, what is that about?

"90f4cfc4-a63b-4d78-b2b8-7372b7fca102"

[tight bluff]

its a unique number

or UUID

[unique stone]

Ah yeah but I mean, what's generating that? Suno songs already have a UUID

[polar spade]

It's a Schlager song, so it needs a Schlager ID. 😁

[cloud shore]

Is this being done with a Persona?

[tight bluff]

no

[tight bluff]

its a internal tool which creates the ids.
We noticed the problem between the night 1-2 november. Before that the suno works perfectly for us

[tight bluff]

nooo XD its not a Schlager

[cloud shore]

I see that the song is cut off at the 1:59 mark. I have seen a few other's also have this problem. I have not heard if they solved it though.

[charred merlin]

Do you recall if it was in this discord?

[tight bluff]

nah private message with a suno user

[polar spade]

I was being a little tongue-in-cheek for the generic downbeat feel of the song. My apologies for poking a bit of fun.

[unique stone]

When you say "we"/"us", are you not just 1 user making use of Suno?

[tight bluff]

me generating songs and i have a partner^^ thats "us"

[polar spade]

I like playing these style songs. Gives me that ABBA feel.

[charred merlin]

not a worry -- I like to connect with "sunovators" to try new ideas, so I wasn't sure if it was something that I'd missed in tech-chat when I was off attempting at a real life 🤣

[cloud shore]

@tight bluff have you tried extending from 01:55, to see if it will fill in the missing lyrics?

[tight bluff]

but we have to use credits to do that? This is ewww for one song ^^

[tight bluff]

^^

[cloud shore]

Does it generate only 2 minute's per generation?

[cloud shore]

Do free user's get time-stuck at 2 minutes?

[charred merlin]

Using the same prompt I was able to get 2 generations that are within v3.5's normal range -- if you find that it continuously gives you short generations, the first thing I'd suggest is approaching the prompt from a different method.

If you're finding that EVERY song comes out shorter, you'd need to contact [e-mail] so they can look into the matter, because that shouldn't be the case 🙏

[tight bluff]

what do you mean?

[tight bluff]

yea, we are generating these text like 50 times (same text except some names) and somethimes 4 min sometimes 1.5 with broken lyrics. Extremly since 1/2 november

[unique stone]

@tight bluff is this a new Suno account you have created?

[tight bluff]

yes, company account

[dense ermine]

i cant make a song it gives me a error every time

[unique stone]

What's a company account?

[tight bluff]

means that we use this account as a company

[cloud shore]

Did you create this account in Suno, directly? (suno.com)

[dense ermine]

any one an idea

[cloud shore]

What's the error message?

[tight bluff]

yes

[cloud shore]

Is it a free account?

[dense ermine]

system error

[charred merlin]

Have you reached out to [e-mail] yet?

Regardless if it's a free or paid account, they are the ONLY official support channel.

[tight bluff]

yes, i wrote an email a few days ago

okay, thank you 🙂

[charred merlin]

Okay I gotcha -- they're on EST business hours, so if you wrote the email on Friday night, it's likely that they're only JUST getting the chance to see it now 😅

[tight bluff]

ok, i will wait ^^

[charred merlin]

Hi @dense ermine -- this is a vague error that simply means "something has gone wrong", but the team has not coded an error message that explicitly describes that the problem was.

You'll need to contact [e-mail] if you want to talk to the only people who actually know what they're talking about & how to help, as nobody on the discord channel is able to provide assistance on this matter 😦 🙏

[charred merlin]

#📚┃suno-school message

🌊

[cloud shore]

I've generated a few times in the past few minutes with no system error.

[fallow depot]

Yo, anyone knows how to properly get a rewind sound in songs?

[charred merlin]

I love the glitchy-rewind effect, and haven't found anything that works reliably through prompting alone, but if you add an effect like that into a suno clip & reupload it back onto the site, Suno can easily replicate that sound -- sort of a "Cheat" for show suno what you want 🙂

https://www.bandlab.com/studio has an effect library built in

[little cargo]

i have a question.... my soundclash remix has dissapeard from my main song list in my profile.... was the songs removed or did something change?

[fallow depot]

aight, thanks for the reply. will give it a gen or two with sampling styles and if that doesn't work I'll try your cheaty method 😉

[little cargo]

nm i figured it out

[charred merlin]

SunoScoundrels4Lyfe

🤜🤛

[sonic igloo]

2 minutes limit on extend?

[charred merlin]

fighter of the nightman -- you've finally arrived

https://tenor.com/view/reaction-happy-dancing-always-sunny-excited-gif-5410501

[charred merlin]

Yes, 1m for v3, 2m for v3.5

[sonic igloo]

Is it possible to continue through a persona? But it doesn't work for me. I've already used up almost all the personas.

for 1 song

[charred merlin]

afaik, you can only use personas in the initial generation, but you can extend normally after that -- extensions use the segment they're extending from as context, so if there's a recognizable beat & melody, suno should have no problem extending from it.

After the free songs, you use credits for them like normal 💪

[sonic igloo]

I understand that the persona is made to sing the same song, but with different lyrics. Right? But he just sings the same song with the same lyrics.
After the 10th try, he changes a couple of phrases.

[charred merlin]

That's the intention, yeah -- for the songs that blatantly don't respect the requested lyrics, you can flag them with ... > Report > Bug > Doesn't Follow Lyrics to help train that kind of behaviour out of future models or to help refine the Personas feature, as well as giving thumbs down to songs that turn out that way -- won't immediately impact the results, but it's harder for the system to improve if it doesn't know WHAT it's doing wrong.

I had hit-or-miss feelings on personas until I got a bit more experimental with how I created them -- at first I made a persona from a song that starts off with soft hip hop vocals and ends up yelling -- all of the persona tracks had the first vocalist or the instrumentals from the beginning and I was getting fed up.

So I cropped JUST the angry vocals as their own clip and made a persona from that -- GOLD.

[sonic igloo]

then it turns out that he is still a child?

@charred merlin

two characters would be cool.

[charred merlin]

he is still a child
Meaning "personas is still in a very early stage", yes -- otherwise I'm not sure what you're talking about.

[fallow depot]

sorry for bothering, but like. when you use uploaded audio, is there a way to recall it in the song?

[sonic igloo]

yes. i mean that.
i just did an extend of a song that a persona song. and he gave out a 2 minute song.

[charred merlin]

Not a bother at all!

I'm not sure if there's a way to explicitly prompt something again, beyond how you'd normally attempt to make it reoccur -- it analyzes the patterns & waveforms that make up the specific sounds & tries to continue them, so having enough data in order for it to reasonably predict how to continue matters a lot

So if your audio upload includes a distinct sound effect, you're likely to have it repeated naturally, as well as Suno possibly having an easier time associating that noise with some prompts -- however, if you're asking it to recall a shorter segment that it might not have been able to isolate as cleanly, it's like asking someone to isolate someone's voice in a crowded room where other noise sources are giving a much clearly, easier to follow sound that might make it hard to focus on the background details in a way that allows you to recall them as easily.

May be a weak analogy 🤷‍♂️

[charred merlin]

Yes, v3.5 is only capable of 2m extensions -- but you can extend from that extension, or use "get whole song" to piece the 2 parts together (and then extend from that full song... we've entered the rabbit hole now)

[sonic igloo]

each exted changes the song. and it turns out completely different. if the persona is at an early stage, and the community should promote it. why limit it?

[charred merlin]

each extend changes the song
I guess we need to agree to disagree? Unless I'm intentionally changing the style prompts, Extend works a charm for continuing the audio it's given.

Unless you're extending from extensions over and over again without bothering to use "Get Whole Song" at any point -- yeah, that will mess with the audio quality, but carpenters and tools and whatnot

[sonic igloo]

is this your project?

[charred merlin]

Not a Suno dev, no -- just a traditional musician who's pivoted to using Suno as a new form instrument -- Extensions are probably my most commonly used feature, and I've released a full-length album that I'm very pleased with. I think there's only a single song of which was created as a "one-shot generation" without the use of extends.

[sonic igloo]

Thank you for your help. Give me a link so I can listen to your songs?

[stark badge]

so is all new features permanently behind paywall now? something as basic as cropping should be available to all.

[tepid narwhal]

Yeah ive noticed those new features havent been shifting to free tier like usual

[hollow hare]

It costs computational power to do those things and that does not come cheap when you think about how large the userbase is.
Besides, that is only a factor if you want to make a song available on Suno. You can download and edit the song in any audio editor otherwise for cropping, trimming, etc.

[charred merlin]

You're very welcome! Since Suno is such an interpretive tool, there's no clearly defined "learning curve" -- so I'm happy to help out whenever I can because I learn just as much from experimenting and familiarizing myself with the platform 🙂

I have a link on my discord bio if people want to check out the more personal stuff I release, but I have a couple silly tracks available on suno.com with the same username, "theharber." 🙏

[stark badge]

maybe it could be mitigated by adding credit cost to the free tier cropping so user can decide to generate a song or crop. why not add same to upload feature. now free tier can upload as much as they want. i would think that costs more.

[hollow hare]

I am honestly surprised about the audio upload feature bring available to free users in the first place. As to why they did not do what you implemented 🤷
You can suggest that to #💡┃suggestions-web - it likely will interest the devs enough to do just that imo

[tiny falcon]

I think suno need an option to allow user to send credits to other. That would be nice. I would love to help someone who might run out of credit and trying to finish their last song before the new monthly cycle end.

[charred merlin]

I personally think that making "mind-to-music" generation available to all for free is a pretty awesome step in the first place -- complaining about access to paid features feels like trick-or-treating and complaining that they didn't give enough candy.

I mean -- sure, full bars are nicer than fun-sized, but... free candy, dude.

[cloud shore]

Perhaps as a 'upload 2 times with credits' or a testing phase for free accounts maybe. But, being locked behind a paywall makes sense in that it allows paying users to upload they're own content (That they have the rights to, because legal things) and use suno. Where as a free account should be seeing if they like to use suno in and of it's self before using the upload feature.

[tiny falcon]

Yeah man. beggars can't be choosers. It cost money to run suno server and it's not cheap

[graceful condor]

I can’t have candy #bloodsugar

[compact raft]

I need help. Every song I create has a SUNO voice ad, but I am a paid service user. I can't use my points now because every song I create will shorten the overall time.

https://suno.com/song/ef49b419-7afb-4997-a5b4-e5819543685a

[neat nexus]

Thanks for reporting this. I will pass it on to the relevent parties.

[compact raft]

Thank you so much！！！！

[obtuse night]

hi, is the track we created in Suno already mastered? or do we need a mastering process on the songs?

[untold briar]

It's not mastered in the traditional sense, but Suno tries to create well balanced tracks right from the start. If you think your track doesn't quite sound right, before trying anything too fancy with external mastering services, you can try using the Suno Cover feature to create a nearly identical version of the track that sounds just a bit better, cleaner, or more balanced.

In the workflows https://discord.com/channels/1069381916492562582/1252267444773326848 channel some users have gone into specific mastering recommendations, from free online websites, to detailed DAW workflow

[obtuse night]

thank you, very detailed! much appreciated!!

[worn flicker]

What in the world? Have you been smoking the AI again?

I'm playing...good analogy

One tweak. Audio doesn't use the noise first Generative nature. It uses more of a midi style, but only in the way that it starts with a note and infers the next.

So more like LLMs

[tiny falcon]

@worn flicker Thanks for the tips. I test some prompt out and it's pretty good.
https://suno.com/song/88dc7a78-8ba2-40cb-aec7-f60e182ceca4

[worn flicker]

I am going to listen, but one thing...lol. I said very specifically that the top part of the song I showed you doesn't work...haha. I must not have been clear because two of you got confused.

The definitions part. Just the structure

[charred merlin]

Are you saying that Suno DOESN'T create songs based on audio noise?

[worn flicker]

I am.

[cloud shore]

You were clear. I saw 'wrong one' and skipped the text.

[tiny falcon]

Lol. I thought the symbol suppose to define the variable for the notes like in programing language, you define it and then call it. lol

[charred merlin]

You're suggesting that Suno constructs the songs 1 instrument at a time, akin to building a song by Midi? @worn flicker

[worn flicker]

That is why I tried it. But no go

[worn flicker]

I am saying that audio generative AI works more like an LLM where it places a note and infers the next most likely

[tiny falcon]

Oh. lol. I am not sure if it works in my case, but it does sound better than putting the text description inside the [ ]

[charred merlin]

I'm going to agree to disagree, but I'm not going to accuse you of smoking AI 😉

[worn flicker]

Look it up and come back to let me know

[tiny falcon]

I don't know about smoking AI. but the song sound good when you smoking, but the next monrning you listen again, it sounds terrible.

[charred merlin]

Nah -- not really digging the vibe, but we're cool to have differing opinions on how it's made, since Suno hasn't revealed any of that.

Are you basing this on having used similar tools like Whisper, Bark, or Stable Audio Open?

[worn flicker]

Did you listen to my song Midnight music?

[worn flicker]

I am sorry boss, I did not mean it to offend

[tiny falcon]

no?

[untold briar]

I think it depends on the specifics of how things work under the hood, but in Suno at least, it seems to be produced all at once, rather than in pieces or tracks that can be indivually adjusted as a human user would typically do in a DAW. But even on that, features like Cover kind of break the mold and allow for other possibilities I think.

[worn flicker]

I was speaking very lightly. I promise I only know what I know, and Suno uses a combo of generative procedures. It is just that those from all I know do not start with noise. If you think about it, that is the opposite of what you want in music or audio in general.

[polar spade]

My only job here is to hold the ladder and keep the pizza hot. I'm good. No worries.

[charred merlin]

You also don't want static in your image-gen, but you'd be surprised with how they start off looking before you see the end result 🤷‍♂️

https://developer-blogs.nvidia.com/wp-content/uploads/2024/07/diffusion-model-building.gif

[worn flicker]

I went ahead and looked it up. My logic was mistaken. I had read about the use of spectrum analysis that is a newer type of generation. My search into future went backwards. Ya'll are correct, current models to start with noise

I do tons of diffusion. Trust me I see it at work.

[charred merlin]

It's okay -- none of us are experts here

We're all learning together, and that's why I love it here

[worn flicker]

The difference between us and many others is admitting mistakes...lol...sad but true

Sad But Truuuuueeeee....here is to you Metallica

Yes you may, I am sorry for derailing it.

[charred merlin]

I know you’re being silly, but I appreciate your input in tech-chat — just a bunch of weirdos in lab coats and I love y’all

[polar spade]

🤣

[obtuse drift]

Anybody know what’s going on with persona’s? Seems like all I can do is just make the same song over and over.

[worn flicker]

Hey, OK so I knew I read about it somewhere. I went and found where I had read that. Unless I am reading incorrect data then Suno uses Bark combined with other methods, and that from everything I can tell does not begin with noise. It uses latent space. If you want to read about it you can. I promise this is not a know it all thing. This is an accuracy for my own knowledge thing, AI is the field I am moving toward. https://huggingface.co/docs/transformers/model_doc/bark

[copper cloud]

Not what I was getting at but okay

[polar spade]

Hey. I like this analogy thing. This is fun.

[worn flicker]

Now that is a good analogy

[polar spade]

My man. I got a cookie!!

[worn flicker]

Damn you...that is accurate...I had to ask for one...bleh

🤣

I wasn't involved with the whole copyright talk, was that about someone seeing if their material is being used or if they were able to copyright the music they made on Suno

I see. I was just wondering. You feeling deep today? You have some poetic things you are saying maybe you should write some lyrics today.

I was just wanting to know for future reference. Thanks

[dense copper]

Does Suno have an API that we can use for our Discord Apps?

[worn flicker]

No API for Suno

You can follow some channels, but asking about APIs for Suno, I will assume you definitely know that already...lol

[dense copper]

I just wanted to play my liked music on my bot tbh

I can do that with streaming downloaded tracks but that's requiring to play local file so...

[worn flicker]

a web hook maybe?

[dense copper]

It has nothing to do with webhook

If I can't access to SunoAPI, the whole system in my mind is just has to be done manually then.

[worn flicker]

I see, I was thinking you just wanted to have the option to play your stuff directly. Hence the ? Lol

Sort yourself out App

[dense copper]

Yeah, filtering liked songs and playing directly or downloading to local and streaming it.

[worn flicker]

I see. Yeah there has been a decent amount of things around an API for Suno. Not from your point but from others being cutoff of an unofficial unauthorized one. That is the only reason I know there isn't one.

[dense copper]

I came here after checking unofficial ones, it is not what I am looking too

It's for generating songs while my purpose is different

[worn flicker]

I wasn't implying that at all. That's why I said not from your point...I should have said not for your reason

[dense copper]

Thanks.

I hope they will work on SunoAPI soon though...

It will be a good news for developers ❤️

[worn flicker]

Being headed that direction myself I agree 100%

[wide grotto]

I use a persona but along with the song style and voice, the lyrics are also copied, what's the point??? 🤔

[polar spade]

You using an older song, or typing in your lyrics before choosing a persona?

[deep ginkgo]

I use chat gpt for lyrics

Or my own

[wide grotto]

persona is only for new songs??

[polar spade]

I'd have to try again. But then, it would work if you chose the persona first, then paste in your lyrics before clicking CREATE, yea?

[wide grotto]

Enters new song lyrics, then adds a persona

[wide grotto]

ok, I'll try

[polar spade]

I can click on my persona... then go to an old song and click "reuse prompts, it will fill in my old prompt and old lyrics.

[deep ginkgo]

Yep

[polar spade]

Or I can then click persona once more and let it repaste the persona's prompts back in styles

[deep ginkgo]

I reuse the persona for different genres. If the goal is to make an album and not sound like it’s AI, you want the same voice. That was my read on it anyway

[wide grotto]

I tried now, after clicking on the persona, I deleted the old text and added the new one. It then created a song for me with the old lyrics 😕

[deep ginkgo]

I use persona for voice only.

[worn flicker]

Same

[polar spade]

Whichever styles or lyrics you want, in basic terms, click "reuse prompt" last or paste in last.

[worn flicker]

Persona has the song it originated from riding under the song you are making, so if you don't have large changes then it will poke through many times.

That is how I do it.

[polar spade]

You won't be covering it with a new voice.In other words

[tacit haven]

can you upload v1 songs from chirp to audio upload to continue without problems?

[untold briar]

Not perfectly, but it usually works fairly well. You won't be able to toggle the song Public if it has vocals, like any upload though.

[low pond]

just 2 minutes of silence 🤷‍♂️
https://suno.com/song/178d5078-01e5-47eb-a20d-a3619a18ad06
https://suno.com/song/15b316f7-cc54-4b7e-a345-163c4583e51d

[tacit haven]

i wanted to ask if theres like any trouble with it

so like copyright or other stuff

[untold briar]

If you generated the V1 songs on a paid account, you have all the rights to do what you want with them.

[tacit haven]

i didnt

[charred merlin]

🤷‍♂️ we can only upload audio that we exclusively own the rights to

[tacit haven]

so we cant upload the audio we generated in chirp v1?

thats right?

[charred merlin]

Like Jon said -- if you made the v1 songs on a paid account, then you own the content & would be allowed to.

[tacit haven]

so basically no

ok

[near tangle]

is there a proper way to designate when a voice should be male or female for following lyrics? I can understand not being able to choose between two different voices of the same gender, but is this possible in v3.5?

[gritty narwhal]

Ive found best luck with this.

Line
Line
Line
Line
[male voice]

not perfect but it works better than putting it at the top, from my experience.

[near tangle]

hrrm

[gritty narwhal]

also depends on your tags

Metal will give you a male as its most commonly sung by men and hense the sample come from that, Celtic - usually female as It seems to pull heavily from Celtic women.

[near tangle]

I can get women for metal

[gritty narwhal]

Yeah, me too, but ive noticed that it leans heavily on men, also depends on the specific sub-metal you choose, (heavy, black, death, doom, etc...)

I work with Doom metal alot so I get more dudes than gals

[near tangle]

I work with either Symphonic or Power Metal

Symphonic is where I have the female voices most

[gritty narwhal]

Yeah, those have a good amount of women in them. I also find it helpful to name them like so.

Thor (heavy bravado):
Line
Line
Line
Line
[male voice]

[tame venture]

Whoa buddy, cool it with the technical jargon

Jk

[worn flicker]

Yea man using words like help and lines, you think you're better than us?....😎

[pine dagger]

i do it with Prompt "BothGender" and say [Verse 1 | Manuel (Male Vocal)] - [Verse 2 | Anka (Female Vocal)] so its work most time good

[gritty narwhal]

what steps do you all take to avoid ghosting? I get some frequencies in my songs that are horrible, and noise extractors have a hard time with them.

[heady shoal]

Is it better to upload songs mpv4 or mp3?

[worn flicker]

Mp3 is more common

[heady shoal]

@worn flicker ★-(ゝThank You!ゝ)-☆

[worn flicker]

There will always be the Ghosting in the instrumentals. Only way around that would be to generate separate tracks and that just isn't how Suno works.

[grave stump]

what is the best way to get suno to make a quick ending to a song?

[vapid nymph]

Have you tried "abrupt end" in the style prompts?

[grave stump]

hmm

[near tangle]

I got it to work with [Choral Harmonies] for the duet parts, but the male and female parts aren't perfect yet

[charred merlin]

.mp4 is video 🙏

[charred merlin]

Following Pox Toxic's suggestion, create an extension point just before you want the song to end & get rid of all lyrics & style prompts except for "end" -- v2/v3 seem more willing to end the song regardless of whether or not it would fit into "the optimal song structure" as it sees it

[pine dagger]

yea sadly you need to waste few credits

[rigid mortar]

Trying to figure out how to do a portion of a song where it's just singing with let's say the drum behind them or acapella. If I use the acapella word in Brackets it doesn't work so any other suggestions on this?

[near tangle]

how often do credits come back?

I used mine up at 3 and it's only been 5 hours

[untold briar]

Once a day. Same time every day. I forget exactly what time refresh, I think it was in a few hours from now, someone else will know.

[near tangle]

I just got mine

so it must be 8pm EST

or maybe earlier?

[untold briar]

8:00 sounds right

[lofty merlin]

Are personas more limited in the duration they can generate? When I use a Persona, the generations keep getting cut off at 3:23 instead of 4:00. Is that expected, or a bug?

[untold briar]

It's expected that generations with some new features like Persona and Cover have less than 4:00 maximum length, yes. You could still suggest longer tracks as a suggestion, but at the moment, not a bug.

[lofty merlin]

Ah, okay then. I was hoping otherwise, but that's fine. Thanks!

[charred merlin]

/refresh

[flint haven]

Free Credits Refresh

Refreshes at 0000 UTC+0 (<t:1713225600:t> in your localtime)

[charred merlin]

Monthly subscription credits renew at the same hour that you subscribed initially — originally became a member at 3am? That’s when they refresh

[wanton copper]

Will Persona feature free?

[hot trout]

Has anyone else had Personas suddenly start acting REALLY weird? I made a whole album's worth of amazing songs using my Persona, and I was so happy with it... until I tried remaking the song he was based off of. It gave me a good version, but it had kind of a bland melody for the chorus, which of course would not do. So I started making a few more... and I noticed something. All of the forty or so attempts since have had the same verse melody, or chorus melody, or BOTH, as the original track. Even after changing the style guide. Even purposely trying to break it. Like, swapping genres and asking for female vocals and stuff. And sometimes, it's not even the same melody/ies... sometimes, it's the exact same !@#! song.

[soft gyro]

???? is there a reason why the prompt is limited to just 120 characters ? It's pretty little !

[obtuse night]

anybody getting "An error occurred.
A system error occurred." when extending songs ? i got one since 10 hrs ago and still

[soft gyro]

with Cover is the same. Almost all of these new functions are very moody / picky. One song they work fine, the next song they do only what they want... cover sometimes make 50 songs in the row the same. So it kinds of stuck. it comes to my mind now that maybe is better to do do one of these things:

1. wait some hours
2. log out, shut down the browser or even the pc, start again, log in again, try again.

[charred merlin]

It's always been set there, I'd have to speculate that they noticed a higher rate of hallucinations or prompts being ignored when anything over 120 characters were used -- only so many "tokens" that an AI can juggle at once depending on what it's made for https://platform.openai.com/tokenizer

That said, adding extra styles in the form of metatags in the lyrics section is pretty common, but common enough that most/all of them get ignored when a lot of them are stacked.

[soft gyro]

except for the case you started the sub in the summertime and you are now in th winter time. It happened to me that last sub it erased all my old left credits and gave me 2500 credits one hour before, so the summertime hour.

[hot trout]

There are tricks that you can use to cheat it, but they tend to lead to complications, so you have to be willing to expend a lot of credits on extensions/regenerations if you want to make the most of those tricks. Then again, you'd be surprised how far that 120 will get you, even played straight, if you get creative.

[charred merlin]

Oooh -- good call, I should note that I'm not official support by ANY means, and https://help.suno.com & [e-mail] are WAY better resources for seeking help regarding billing stuff -- lots of edge-case stuff that the community might not be aware of

[soft gyro]

I am very sure the AI does not hallucinate at all if you place more tags there, cause there is an interpreter at the top that picks up randomly what you put and make real commands for the model out of it. This is how Udio works and how all are working. Udio has no limitation. In Suno you have to change all the time the tags.

In Udio you can put a lot and let the random pick it, which is better.

The hallucinations have a very different cause !

[soft gyro]

which kind of tricks? abbreviations?

do you know that Ai is so intelligent that if you write smth wrong or with missing chars it will still understand everything ? So you can trick a lot

[hot trout]

Hang on, lemme grab you the link that explains some of the method...

[hot trout]

https://sunoaiwiki.com/tips/2024-05-28-using-detailed-prompts-in-suno-v35-lyrics/

[soft gyro]

Oh man! Who the heck made this ugly Knowledge Base with a WHITE background ???

[hot trout]

I don't understand why ANYONE uses white backgrounds anymore; it is what it is.

[soft gyro]

this 120 chars LIMIT is a LIMIT to creativity too !

but you can add also tags in the lyrics window

[charred merlin]

That's likely overkill, though -- SO much time spent proving the majority of this gets ignored, or is largely interpretive & subjective so it's hard to determine what ended up in the "final mix"

[hot trout]

Yeah, do NOT use all of that. But a few elements, used juidciously, can really make a song better and give you finer control.

I find I have the best luck using a structure like that, but only focusing on genre, vocal style, and mood.

[robust rune]

website is down again

[hot trout]

You can also try giving it a key, or even a chord progression, but only the great god Suno knows when and if the great god Suno will choose to honor your puny mortal prayers, LOL.

[unique stone]

Sites need a mandatory flashbang warning for this kind of thing as well as those annoying cookie acceptance popups 😄

[soft gyro]

these minimal tags define a song:

• pace
• mood
• genre
• type of voice
• extra voice elements (duet, choir)
• some instruments
• some properties of the instruments
• key
• year
• sound effects
• anything else ?

This is already 73 chars: slow emotional romantic song, warm female voice, choir; grand orchestra;

you dont need space after ; or , - so you can save some chars

[hot trout]

You're using extraneous words and symbols there. Taking up characters, AND making it less likely to understand you. For example, I'd rewrite that as "slow, romantic, emotive female vocals, orchestral, choral". Suno tends to do better with short, choppy, declarative tags.

For pace, don't waste spce in your style prompt; use in-line tags for that. Want a slow, lilting verse? Use a [SLOW TEMPO] tag after declaring the verse in the Lyrics field, and nine times out of ten, it will take.

[unique stone]

As extends already take the previous part's context into consideration, it can help to space things out across multiple parts as well. So I'd start with an instrumental, prompt for the styles and instrumentation I want. See how that sounds. If I want vocals, style those in part 2 (extending early if I want a short intro.) If I'm making an adventurous/progressive piece I'll only prompt for the things I want to change. I've got a 7 minute long song I've been adding to where each part is just like 2 or 3 tags in the style prompt to add or change things.

[soft gyro]

yeah, but extend also deteriorates the sound...if you do it at the intro already....even worse

[charred merlin]

Hey dude -- I know we've shared songs before, I really enjoyed some of the stuff you've made on Suno and outside -- I trust that you'd give me an honest opinion

I use Extend a LOT in my songs -- If I show you one of my songs that used a lot of them, would you let me know if you feel like it's notably "bad" for having used them?

[soft gyro]

after 3 weeks of break with Suno I made a good song again with Suno, a song which I imported from Udio.
It is much faster to make a somg with Suno and Suno sings MUCH better too. Udio started great but failed after...

[charred merlin]

I completely understand how using Extend over and over without using the "get whole song" to piece everything together for context can make things sound progressively worse -- and I know that Extend will continue any quirks or negative audio present in the original segment (so mastering & reuploading the original 2m segment can be SUPER powerful), but I genuinely don't think that Extend worsens the audio quality inherently

[unique stone]

If you extend later in the part, it is noisier. Extending earlier on, and often is expensive but I find it helps.

[soft gyro]

I dont pretend to have the best ears, but after extend I notice almost always the difference

a way to fix the extend bad quality is to make a cover of the song and have the luck that you get a copy and it sings also the copy

but specially when you need y copy extend makes a new version and viceversa

[stiff plover]

👋 I'm diving into suno after a big hiatus so bear with me if this is something known.

Once in a generation, it creates an absolute BANGER in terms of quality of the vocals and the music, and then I'm unable to reproduce the fidelity hundreds of generations later. Is this something that can be controlled to some degree? I have examples.

[soft gyro]

I just used the persona functions and it gets me the same voice with the new style I type... so it is useful to make songs with the same voice.
but I have to listen carefully to the voice now to check whether is indeed the same voice...

I made thousand of songs and I only remember of very few songs with really good quality and I guess all of them had no extension.
But sometimes you get only bad quality... the AI seems to be influenced by the stars and planets... 🙂

[charred merlin]

OH MAN -- are you a paid subscriber or on a free account?

[stiff plover]

Paid

[charred merlin]

Grabby just mentioned the persona thing, but it's remarkable how effective it can be when used with the right song

[stiff plover]

Even with persona, I get very smiliar style but the quality is decreased. I created a persona from the song that turned out too good

[charred merlin]

1 sec -- I described my experience with it when trying to get a specific tone out of a persona #📚┃suno-school message

If there's a section that REALLY highlights the style that you want to replicate, crop out JUST that section and make a persona from that -- if you have a 3m song, but there's a 15 second clip that perfectly showcases everything about the song that you love, don't leave it up to Suno's interpretation to guess what you want, ONLY show it that 15 seconds and make a persona from that 🤘 When I first made a persona from the 2 rapper song, it was hit or miss, maybe 1 out of 10 -- isolating my preferred vocalist and removing the extra instrumentals increased the success rate phenomenally

[soft gyro]

I made some great songs now one after the other with this prompt, try it too ! And the audio quality is really good now!
new Wave Synthwave 808s DX7 High-Hats Moog Synth, Percussion; drum machines; contemporary sound; male voice
( I even forgot to put melodic)

wow, great songs !

long time I did not get anymore good songs and good quality

but I used persona !

[charred merlin]

https://tenor.com/view/the-fonz-gif-18114253

Grabby just kicking the jukebox, like "play me a masterpiece, sweetheart" 🤘

[soft gyro]

could it be that it is because I used personas? I forgot to take it away

[stiff plover]

I'll give this a shot. So I made a persona from the song, then literally tried to regenerate the same song. It did follow pretty much same exact pattern but with severely reduced quality

[charred merlin]

Could depend on what persona you'd used, but seriously... 9/10 of my most recently "liked" songs are from personas... I think they might have made it a LOT better from the Alpha release to Beta 👀 🥷

[soft gyro]

I used also better audio quality and much better compositions when you user Cover... so maybe it is different when using these 2 functions

[lucid rune]

Anyone seeing this?

[soft gyro]

I have already 4 full concerts of 2 hours of the VERY best music I made...I did not sort it albums but in concerts 🙂
And I have more stuff to finish for about 2 more hours..this is my yield since may, but there were months I did not
have time to make music...

[lucid rune]

like when they try to generate not literally that image upload

[charred merlin]

Have you tried changing the style prompt / lyrics to see if it happens with every song?

[lucid rune]

can hurt to troubleshoot more will try different settings