#🪲┃bugs-web
1 messages · Page 18 of 1
It seems there are two songs that won't extend
And I really want to extend them because they are really good
Yeah i understand that
Then i hope someone is here soon, who can help you. Seems like a system issue :/
I mean, wouldn't you want to build on this? https://suno.com/song/8f6e6f96-234e-4df4-b5ac-062a8667abee
Just for the lol, i tried extending it too, i get the error too 😄
I broke the algorithm
LMAO
I want to extend from before the high pitched nonsense
I was getting so many of those I gave up. Now playing elsewhere until Suno is fixed.
Why the hell would SUNO, when asked to continue a song from a certain minute, continue the song from another place altogether, plus add something else of his own? I spent over 800 credits stupidly for suno to try to continue the song normally! Why is such a simple banal solution like: continue the song from the right place and repeat the chorus the same way you sang it at the beginning, it just doesn't work! Chorus she constantly forgets how she sings, even though you write the same chorus tags all the time. I'm not asking for too realistic vocals, I'm asking at least for the continuation of the song from the right moment (words), repeating the chorus in the style that she sang.
Dm me where elsewhere is please?
I am very upset, I just finally connected the subscription to suno, I thought I would write songs normally now, but as a result, when the subscription is connected, suno works much worse when it is not connected.
I was already doing a normal song for free for 20-50 credits, and now 400-800 credits go with a subscription, and not because I don't like the result or it sounds crooked, because the neural network can't stupidly continue the text that is prescribed to it, it starts inventing text out of thin air. If the neural network cannot work normally in the continuation of the songs, then make it sing for 4 minutes instead of 2, and put the cost of credits at 20 and not 10 for 2 results. Yes, I will be ready to pay at least 50 credits so as not to spoil my mood with a crooked continuation. I want the neural network to live in its song and continue as it planned at the beginning, and not so that when it continues, it forgets everything at all.
And for those who need a change of rhythm and other things, make special tags for them for the neural network.
I'm really very upset with Suno's neural network.
After yesterday's update, an outro is produced very often in the 1st part of the song, which leads to a pause in the song. Parts are also produced without sound, without corresponding credit. In total, not even 1 song could be produced from the daily ration.
@ornate cape i totally recognize those problems, but I think we have to consider that this technology may not work quite the way we assume. Getting precise chronological control might not be quite as straightforward as we think. Also, adding new tags would probably mean going back to the huge data set, manually adding those millions of tags, and then retraining the whole thing, which is costly. My assumption is that Suno is a "diffusion model" and works somewhat like Stable Diffusion which also lacks precise control.
That said, i'm only guessing how Suno works. There's a severe lack of behind-the-scenes info about the technological side of Suno.
I just really would like Suno to work in her contextual form of one song and repeat her own choruses as she sang them at the beginning of the song and for her to continue singing when it happens on extend
Yeah, that totally makes sense.
There are other ways to get more fine grained control over vocals though. Using Suno to make instrumentals, and then adding a Synthesizer-V on top using a DAW is worth a mention. But be warned, it's a lot of work. At least, that way you will be able to adjust each and every syllable.
Diffusion models are great for random creativity, whereas Synthesizer-V, Vocaloid and similar less AI driven technologies leave more choices to the user, but require more work and music knowledge to sound natural.
And as you can tell from 99% of all vocaloid music out there, most people don't have the skills or motivation to achieve realism when using Vocaloid.
This is weird. I'm generating more songs and almost out of credits. It sounds like it just skips beats at the beginning and there is no intro. It's doing this to all of my generations. I tried different prompts and lyrics, and still no fix. Am I the only one facing this? https://suno.com/song/46ec9ccf-755e-4456-99b3-452d60c201e7
So I believe it's important when reporting "bugs" to make sure that they are in fact bugs, and not limitations in the technology. The weird reverb effect on Suno v3's vocals I think is safe to call a "bug" considering v3alpha did not have this problem.
Simply because we've seen it work. And now it doesn't. So it's a regression.
you have to make new lyrics for each part
or do you have lyrics in the box and it doesnt sing them?
@lusty pollen but SUNO does songs out the lyrics , just doesn'
doesnt write it down
you have to generate new lyrics for each part. i usually just copy the main chorus of the first part and then let suno write new lyrics and replace the chorus
@lusty pollen OOPS...I thought SUNO will write it automatically..... so I just empty it.
but I think that even I empty it, suno should write the lyrics it sings.... it's hard to write out what it's singing
that could be hallucinations. if suno wants to sing ... it does 😄 its gibberish or fragments of the already sung text.
If suno is like bark it is mostly an LLM just trained to spit out encodec tokens instead of words
i've been playing around with bark too. but i don't know what "encodec tokens" are. i only have a vague gut feeling how bark's text-to-speech works, but going from speech only to music puzzles me.
Hey i am unable to access chirp beta channel. Admin can you assist?
Chirp has been retired. The website at https://suno.com is where you make music with Suno now
Okay noted ty
I came back after a week to see if the bug that caused the first few seconds to be cut off after selecting "Instrumental" mode still existed. It still exists. It also exists in custom mode, which was not available before
Sadly V3 really quiet, while V2 results are distorted like crazy, been like this for over a month now: https://suno.com/song/c5929549-437a-4643-89d8-48df64c21fd2
@cinder sparrowFinally, someone from SUNO TEAM responded that they were aware of these issues.😂
Therefore, I really hope that when the next version of the model is updated, the previous version can also be retained for people to choose from, including the SUNO V3Alpha a few days ago.👏
I really only use UDIO to create music recently, but I still look forward to SUNO. UDIO is great at making complex music, but music is not only complicated. I still like SUNO's seemingly simple but emotional music.🤤
May be dumb question, but I made an old song in the beta chanel before tthey had the website going. It's not in my library and at the time I only saved the video. Obviously the beta channels are gone so I can't search for my mention, but I found it my date as chirp bot like. https://cdn.discordapp.com/attachments/1132066051962703992/1186274890114928670/chirp.mp4?ex=664ddc95&is=664c8b15&hm=89233c6d172c24cd1636ea1e920528a40a2d8fac2bdfe7f1f53bca2af2ee6edf& #😍┃v1-songs message Is there any part of this URL that I can paste as the song identifier? So as to replace the identifier in their ural format eg : https://suno.com/song/7f3f3fe3-dd69-4968-9fdb-8b170c6ab482
I've spammed posted this sorry, but it's being ignored in the other active channels, and wasn't sure where else tt ask.
Hey there, I'm certainly no expert in this as I wasn't even around before the website, but as far as I know that's probably impossible, since the website is not really linked to the discord, it probably has its own storage and probably didn't transfer existing songs from the beta
But feel free to correct me (anyone) if I'm wrong
I've experienced some oddity with the site today - unfortunately I don't have any videos of it. Here is an organized list of what I've had:
[1] Earlier today, I was renaming some renders I had - I would rename render A, then go to rename render B, and render A would revert back to its original name. Go back to rename render A and render B would revert back. This seemed to resolve after refreshing the page, but it was still a weird issue where renaming wasn't sticking
[2] There was one or two times today, I went to my Library and the page buttons were greyed out, I could move off the current page of renders I was on. Again, a refresh seemed to resolve it.
[3] This has been the most on-going one. My Create page has been slowly emptying itself over the day. Normally there is a page-worth of renders and older ones drop off as newer ones are created. I had a whole page of renders in Create this morning and as the day as gone on they have been getting removed. Now at the time of this post, I only have the latest 2 renders in there, stuff I created within the last 2 - 3 hours have completely been removed from Create randomly that I didn't downvote or delete myself.
Since several days lot's of songs start directly. Here an example.
https://suno.com/song/a0519418-ef70-482f-b173-2f090860feeb
trance piano melodic-dubstep song. Listen and make your own with Suno.
Hey there. I have experienced the same. Seems like the first 1-2 seconds of a smooth start are missing: https://suno.com/song/e93c324b-4888-4b88-8956-f320b0eb4453
cinematic, epic, space, fresh start after sad times
song. Listen and make your own with Suno.
But great song!
Hey again. I experienced some rather severe distortion/degeneration of sound quality in my song today. Like too many layers overlapping. Or like the mic and instrument cables too loosely connected to the amplifiers: https://suno.com/song/85388bc2-f8b4-490f-be7c-a158590929ee
Pop-rock with motivational and uplifting elements song. Listen and make your own with Suno.
I've been experiencing this to.. It just just directly into the track without any intro; sometimes it drops into a soft building section as like a psuedo intro after 10 - 20 seconds, sometimes it doesn't and just keeps going
I've also been kind of experiencing this too, its like there's a formant fx overtop of it. I've been getting this especially in the vocals
it would be cool if you could post the suno link to a broken generation. its way more easy for the devs to have a look at examples to find the issue.
Here's one example, https://suno.com/song/25883316-9a0a-4972-beca-0ebd7dc5c5f0
Future Garage 2-step half-time melodic atmospheric dystopian broken-jungle-beat lots-of-reverb song. Listen and make your own with Suno.
Here's another with really degraded vocals: https://suno.com/song/a1fa8a89-94db-4a48-a3db-e20eec995965
Future Garage 2-step half-time melodic atmospheric dystopian broken-jungle-beat lots-of-reverb song. Listen and make your own with Suno.
It's been doing this with all of my generations, thank god I know it's not just me.
I've been finding it difficult to generate beyond a part 2 or 3 without the tonal quality being noticably degraded. My current work around is to keep hitting generate and rolling the dice until I get something that sounds like it has a fitting tone/quality, but also an agreeable rhythm as well. 💸
If I follow a link to Suno through Facebook Messenger app on my phone, I'm not logged in, and if I try to log in using a Google Account it says "Access Blocked: Suno's request does not comply with Google's policies" "Suno's request does not comply with Google's "Use secure browsers" policy.
suno is reading [bracketed instructions] as lyrics. It didnt use to do that
Well it was less common in V2, but it has been a thing from Day 1. Day 0 if you count Suno Bark. But in general, keep the [amount of text in brackets] shorter in V3 relative to V2 to avoid it. If you are explicitly prompting an instrumental and you don't want singing of any king, it may help avoid obvious style words like 'vocals' as well as styles or adjectives typically used to describe voices.
This has been happening to me too. I can't get any intros working at all anymore. Here is another example but this has happened on almost all of mine. https://suno.com/song/98dc4fc1-9e1d-44ab-a3b3-c6c9a62b66ff
Punk Rock, Aggressive, Youthful song. Listen and make your own with Suno.
Be sure to change things up instead of just retrying, without changes. Change the style/lyrics slightly between retries, and especially change the Extend From time, even adjusting it backwards 1 second at a time.
The nuclear option is to force it to 'forget' a lot of the music and use a very short clip. But this may hurt long term musical coherence.
vocals are nearly impossible to control anyway
Try extending from an earlier time
also works to stop it from ending the song early
sometimes i will only use 10 or 20 seconds of an extension
Currently the engine can't compile 20 minute tracks - it begins to introduce all sorts of artifacts when splicing the segments together. I'd love to be able to make 60 minute tracks, as goofy as that probably sounds!
will they ever fix the sound ? it all sounds like someone sings through a fan
It used to start failing at around 10 minutes, so 20 is already an improvement. All engines have a limit as I can only assume a bigger duration needs more resources. For example, Udio has a hard limit of 15 minutes from my knowledge. I think Suno should also enforce a hard limit and make it visible in some docs.
of course that's ignoring the fact that it worked fine with the "lost version" ... V3alpha
no doubt you are messing with the sound quality but doing renders after a couple of days, they all have way too much reverb on the vocals and too much top end. Does your head in listening to a whole track on Deep House and impossible to fix in a DAW. Has gone backward over the last few days.
Hello! suno team!I want to give you feedback. Just now someone else's song appeared in my account, and it existed all the time. It didn't disappear until I pressed F5. I clicked the choose bad button, and their song will disappear. And I ask the song to sing in Cantonese, which often suddenly turns into Putonghua, which wastes a lot of my credit.🤖
So I used to create drum and bass tracks with lyrics being very short, maybe 2 or 3 words, now they just get refunded and doesn't happen 😦
So I'm also getting tracks that degrade very badly after one or two extensions. Practically unlistenable 50% into the song. This is very noticeable with rock music 😞
yess all the song the longer it get it sound like the intro of .. 'killing in the name of' from rage againt the machine lol
cant wait to see what v4 will improve v3 is more like a toy right now
example for metallic ringing noise starts from around 40s , intensifies towards the end. https://suno.com/song/8027c4b3-c525-43d2-b8b1-1395eed7cc58
Creepy Halloween, darksynth, electronic drums, german electronica song. Listen and make your own with Suno.
here is another example of a different user... we tried similar styles and he also got the ringing https://suno.com/song/7b4241a6-f3e1-4b83-a0a5-001aafec8968
Creepy Halloween, Somber, sinister, western swing, Trumpets song. Listen and make your own with Suno.
https://suno.com/song/a1fb83c2-ec7c-444e-bb13-fcfca6bdc36f Finally found a solution to the Cantonese problem I asked for haha
Hong Kong Cantonese
smooth kids music song. Listen and make your own with Suno.
And the Cantonese of SUNO is very good.
Did you extend the song at that point, or is that from the initial generation?
these are both first parts
That’s not good… hopefully the issue gets resolved.
New album created here -> https://suno.com/playlist/37a63935-11ef-471b-934b-fe449550e4d6 - link in discord does not show my uploaded Album art 😦
it would seem with the update today that i am now experiencing double scroll bars when viewing a song full page https://streamable.com/3mhhzu
Will take a look, thanks!
Hey er'body! I noticed a change recently I'm bummed about. I was creating lots of authentic, old, retro sounding songs with Suno using descriptors like "1960s, old, retro, vintage, vinyl" and this was working WONDERFULLY to make songs that sounded like old records! And then recently something changed and everything I've been generating is modern sounding...or trying to be modern sounding. Any idea why old styles aren't working any more??
When going to a /song/ page directly (e.g. suno song link from Discord) or refreshing the page while there, the Suno Logo (top) by the credit and the Create button (bottom right) image references breaks.
Likely because these images are using relative paths to the images rather than an absolute path (thus looking for the files in suno.com/song/ path rather than the root path). Should be an easy fix.
Happens also on playlists
We're looking into this, thank you!
Also have a bug can't change my payment plan, I accidentally clicked on the wrong one I wanted monthly and now it can't be changed 😦
The image rendering issues should be fixed, but let us know if you still experience any
If you haven't done so already, contact billing@suno.com
so don't get any solution here
Tried emailing but no response for almost a month :/
@azure hamlet could you include a link to an example of your image rendering issue?
that is the link in question, it must be really long
sorry could you add complete instructions for a reproduction of this bug as well?
ok
without looking I bet this is base64 image url
I generated a new image in the edit details. I selected the new image and saved it. Then I opened the image in another page in order to save the full size version, and I ended up with the link I sent.
Heya - about a month ago I made 3 full length songs. In my library today I can see all 3 songs, but 2 of them play the same track. Not sure what happened, but it seems like one of my songs was overwritten?
ahh so for reference @unreal radish
before we could append _large at some part in the image url to get the "HD" Version. Now the image is linked as base64 image url at least in the case of @azure hamlet where that doesn't work
Once you refresh the page you should be seeing the “real” image URL
no change on refresh, tried opening into another tab again and it autodownloaded the small image instead
So it was a change then
yeah, the download starting was different
wonder what you get with "copy image link" on the refreshed page
it looks like the same really long url
sent
tried it on another song to test and got same
Hi, whenever I go into a playlist to extend a song it brings me to the Create page with the song linked. I then press Create and the Create page minimises to show the generated songs (one generation). I then click back on the Create page and the song is no longer linked, which means I have to go back into the Library and find the song again in the playlist just to do another generation. This is frustrating as I do 4-5 generations per prompt, adjust the prompt and generate another 4-5 and so on. Having to go back into the playlist to find the song each time, makes it a really long process.
I'm on IOS using the web page app on Safari. Thanks!
Is there any chance my views are stuck at that number? It's been a while and I haven't seen the numbers go up even though I shared it with my friends. I'm not sure, but could it be a visual bug? How would I fix it if that's the case? Or is it just that no one played the song again lol.
Thanks for this feedback - we'll review this!
I am the bug 🪲
I have to bring this up because its off putting. When we download for a video, after a couple extensions in the song its still picking up extra lyrics that aren't used rather than it syncing to the voice of the song. If there is any way you can Sync the Lyrics with the voice and cut out extra unused lyrics per extension. I will be extremely satisfied with the videos. unfortunately i do not have the ability to just craft my own video 🤷
This is fixed now 🙌
Hi folks, Suno keeps messing up my lyrics, skipping lyrics, adding things that aren't there and mixing up lines. I have wasted so many credits trying to get the lyrics I want in and the reason it's important is that I am writing them for someone as a gift and they have to be all in and non skipped because they all mean something that I want to say. What can I do about that and my wasted credits? Thank you.
Btw, I am EarthlyBookingAgent486 on Suno
This is happening in the extension.
Tried the new feature, adding custom images to songs and playlists, but they disappears after a while and goes back to original. They are my own created images.
maybe the resolution is too high? i tried 2560x2560 and suno didnt like that. 1024x1024 worked fine
ok tanks mine are 3000x3000 so I'm gonna try 1024x1024
In mobile the place to press to make my lyrics not working, Desktop mod do help and fix the problem, I believe is because the message popup that need to accept, not sure but yes I can't press the place I need write lyrics. Not big problem at all but maybe good to know about
There is a huge bug when you have first extend a track. The extended track will be extended when you Reuse Prompt and want to start a new song.
When I've done all my editing, reusing prompt, extending and changing lyrics to what I find best and finally creates a FULL song, the lyrics in the video isn't what the song is at all. It has a lot more text then my FULL song has. So it looks like crap when I watch the video. Is there a way (or will there be?) to change the lyrics afterwards WITHOUT changing the song (Just the lyrics that shows up in the video) so that it follows the song
It seems like the more you extend a song, the more it tends to become "scratchy". Anyone else notices this?
I'm fine with it as long as it is just the song because it doesn't matter for the music in itself BUT it's a crap that an extended song gets so much EXTRA lyrics in the video
I create my lyrics before generating the song, to minimize extra words.
I doesn't always help because Suno sometimes just skip words if thy're not compatible with the songs tempo/style and often Suno add words in the beginning of the song. I want to edit that for an exact full video
I also want to add that you can’t see the “Extend from” view when you think you create a new song but instead you just extend the song you are reusing the prompt from.
When I add a song from the new search view to a playlist I get the message that "0 clips has been added to "Playlist Name". When I open the song and add it to the playlist it works fine.
I will add one more thing. If I push the new song button, push the clean button and then create a new Song, it will extend the song I just extended… so it has nothing to do with the Resue Prompts. It’s the Extend option that is the bug here….
Every single time I generate lyrics from scratch in Custom Mode, they're almost exactly the same, always about driving in a fast car with neon lights in a futuristic city, regardless of style prompt and song name
It seems that the update had some really bugy “moments”.
The whole app is trash
Which song it has this issue?
If it's always happening when you extend from the same clip, keep trying different Extend From times (and adjusting the lyrics to add back in any lyrics that you removed by moving the time back earlier)
Sometimes you have to back up quite a long way.
The thing I don't like about suno is the 'growl' that is in almost all the song tracks. I suspect this is to 'identify' it's suno being used. But I'm paying so it should go away.
Some of us can hear your 'suno hum'
Even if it's not intentional it's obnoxious.
Trying to extend a song with additional lyrics but it keeps making the extension an outro and it's not following my lyrics at all. Tried 5+ times. Tried starting the extension at different parts of the song, but same thing.
Have you tried [bridge] or [verse] before the lyrics? Sometimes that can work.
I have now wasted a lot of credits trying to generate new songs and Suno thinking I am continuing the last one. Even after hitting clear. I've been reporting this bug but the error isn't listed and no one ever asks me what "Other" is.
@sweet spruce I am experiencing the Extend bug.
So it seems that changing the Album artwork of a song doesnt change anything in terms of a downloadable video, or embedded Music videos. Whats the point of it then? Is it a bug?
I think it's mainly intended to just change how it appears on the website. It would be nice if it re-generated the video when you've chosen the new image though. I guess they might have this planned for a future update (e.g. after v3 was released, we didn't even have lyrics on the videos for a while.)
Yesss I have, I tried [bridge], [verse], [instrumental breakdown] 🥲
guess i'll wait for some more updates on the website and try again later
Suno team. Bug with very short duration still exists. Maybe some tags in lyrics break something. For those songs extending is not possible or if even is, then it extend it for few sec. only. User lose credits for handicapped track. In this case credit(s) should be refunded. Simple checking if first song generation is longer than 1 minute. Ps. 3 songs generation in a row: 32, 33, 20 sec.
Really hoping so. Would be nice to have Suno Embeds with the same artwork we make for our videos.
Should be updated soon
Ill keep my fingers crossed :3 I updated both my parts and the full song with my new Artwork.
Yo! Why do i have a song with another person's profile? What the???
Oh all the songs i just created
who is Profile
GamelanHyperpop517?
refresh the page. its a visual glitch
or this is your username... you can chage it when clicking your profile picture and then edit profile
I tried several times but no changes
Oh? let me see
Ok you r right. Thank you👍
So, I'm sure this has already been noted, but on mobile at least there are issues with continuations
Ok, removed my second bug as that appears to have been fixed on the recent UI update
but yea, if you do a continuation, it works
but then when you go back to create again, it remains in continuation mode, but without the timer
so you can no longer select the point of continue
which btw is another thing on mobile that's ongoing: it resets the timer even previously on every gen, which can be a little annoying as you'll accidentally waste gens at the wrong point if you forget
but I digress, this would be core functionality, as it means in order for me to make variations, I have to scroll back through pages, find the right song, and re-extend every time I want to produce an extension
Thanks in any case! The new update is quite nice overall and the UI feels subjectively better
Thank you for this - we'll look into the extending experience on mobile!
Another bug: Instrumental generates lyrics. 🤦🏾♀️
Should we post glitched out generations here? Do the devs want to know when that happens?
help
it's vocalizing any prompt words that deviate from the expected in the square brackets. Same thing in parentheses. Not good.
Parentheses are supposed to work this way (it is just sung differently, and not meant to be used for prompts.) Prompts in square brackets tend to be sung if they are not understood. For example I had [birds chirping and tweeting] sung.
What prompts in square brackets are you encountering this with? (Might be useful for the devs to see)
That's kind of the point, prompts in the square brackets should not be sung at all let alone verbatim. i.e [tropical Beach Sounds] it will sing it word for word.
the vocal tempos are out of pace with custom lyrics, its struggling to find simple rythems it was doing inn the last version?
last version?
lately, something is wrong with suno .. extending a song with additional parts of lyrics completely changes the style and voice
So if you had a loudness or degradation issue within that minute, it would start becoming emphasized as that is what the AI picked up on and all future extends builds up the issue as long as it exists within the 1 min of continuing.
It also explains why the style can change when you vary the starting point by a few seconds.```
Try changing your continuation point. If you are extending a part 2 song, you will generally get more consistency by combining it with Part 1 first
it's already a combined melody and it doesn't even follow the style from ~1min prior to the continuation point .. out of curiosity, what happens if the clip length before the continuation point is less than 1min? is it not going to generate a stable continuation?
I had zero credits leading up to my subscription renewal, yet when I was billed my account received 9000 credits instead of 10000. Whats going on here?
From what I remember, nothing different happens when extending short clips, it is more of an upper time cutoff.
It is an AI and how it relearn to mimic the style (and what it actually picked up) can be a mystery.
I had the same issue as you a few times in the past, one of them needed me to go back quite a bit to the melody I wanted and redo the lyrics that followed it.
(For one song in particular, where I loved the melody, I simply extended it as an instrumental part with various instrumental tags and used it as an extension part to make various tracks and combined them in a DAW, though this combined song was a bopping instrumental track at this point)
what kind of instrumental tags did you use? can you share? tags are a big mystery for me, some work, some dont. i have seen the official list, but it's too short.
As it is currently, it's totally unusable.
For my general go-to, I use the following:
[Catchy Instrumental]
[Melodic Instrumental]
[Build]
[Drop]
[Catchy instrumental]
[Big Finish]
[Fade out, End]```
As I said, this is more of a general catch-all tags that I use instead of the instrumental switch.
You can be quite specific in what you want, check out what RayanAIR did in their song https://suno.com/song/a98da392-ee0d-4fad-ba9b-937f0c9c5a91
If you did not know about the wiki, see https://www.suno.wiki/faq/metatags/instrumental-tags/
Checking the video generated and seeing what tag brings the beat you want is helpful too. In the video I uploaded, the Latin part of the Latin/Mozart fusion happens at ~40s because of the [Drop]
neurofunk drum and bass song. Listen and make your own with Suno.
You should contact billing@suno.com (I think) about that
Why the fuck does everything I come up with comes out sounding like imagine dragons
Throw some nonsense into the style prompt, strange words, use lots of dashes like oatmeal-anti-jazz-acoustic. Basically mix it up.
That’s exactly what I do I’m afraid
Well then you must have offended the Imagine Dragons god and are under the effects of a lifelong curse
help
usually they dont get stuck for no reason. is the song family friendly? if yes you could post it here so we can see if it gets views
okok
It’s been like this for months. The number of visits isn’t going up, even though several friends have listened to it. It just stays the same.
8 bit retro Rock videogame style, melodical 8 bits song retro sounds. Unique Song song. Listen and make your own with Suno.
I sent it out before. Do you think it might have been banned for spamming it too much on Discord?
Actually, I used to send the song on Discord every few weeks. It's not like I was spamming it repeatedly. I would just share the song again from time to time when the occasion arose, so people who hadn't heard it before could listen to it.
Oh, Wow, I think it's fixed. Thank you so much! xd
yes i just saw that its working when i looked at the page
Thank you so much, I swear the visits were stagnant. But now I see it’s fixed. Thanks a lot!
No hablo ingles pero tengo un atraco al usar la pagina, y es que cuando quiero tener la cancion ya completada no logra fusionar la instrumental que viene ser la segunda parte, solo fusiona la parte 1 y 3 que viene ser con vocales, espero puedan dar una solucion a eso, gracias y buenas noches
i had around 6000 credits and my $30 monthly subscription renewed putting me at 10000 credits. shouldnt it be at 16000?!
Credits do not roll over
It's always been this way
its says right there my credits do not expire. but ive lost 6000 credits
Read again:
"Credits included in subscriptions do not carry over from day to day or month to month."
The credits you get each month do not roll over. You can not 'save' your credits
When your subscription renews it resets it to what you subscribed at. In your (and mine, I have the same sub) case that means it resets to 10,000
If you purchase extra credits
Those do not expire
Which are these ones
so i was better off not renewing my subscription and then holding onto my 6000 credits?
Which then show up like this:
No because those credits would go away
You do not get to keep the credits that are a part of your subscription past your subscription date
You have a month to use those credits
hmm guess i should sub a lower tier then cause its hard to use that many credits
A lot of people do that yeah
In my case, as you can see, I end up using more 
So I have to 'top up'
You can also upgrade from the lower tier in the middle of the month, if you happen to use more than you expect.
kinda bummed but oh well its the best 30 bucks ive spent in my whole life so i cant complain
A bummer you didn't use those credits. You've got the right idea though, just sub on a lower tier next month
For this month though you have 10,000 to play with
Burn em' up! Try weird things
You can always email and explain it was your first month and it wasn't clear how it worked, from what I hear Suno is usually pretty chill with stuff like that, but I can't promise anything.
yeah im tryna make hawaiian steel guitar ganster rap so i guess its getting weird
We have a regression at least for mobile (i didnt checked on desktop) when using the extend.
Before the last suno update,we could click the extend button several times so that we could make several générations at once . It is not possible now.
- The first part can be 2 minutes long. Why is the second and later parts only 1 minute long?
- The rhythm and vocal effect of [Chorus] in the first part are consistent, and it is difficult to keep the rhythm and vocal effect of [Chorus] in the second part consistent with that in the first part. Is there any solution
Been trying to do another metal / rock song for days now wastig tons of credits but no matter what I do it all ends up in reverb hell, sometimes pretty much ignoring any style directions I try to give. Very frustating. Small example of just right now: https://suno.com/song/e3994971-cb9b-4abc-9c9f-66a4efc85b13
Hard Rock, Stomp, Power bass, Clean male vocals, 136 BPM, heavy guitar song. Listen and make your own with Suno.
The overall bad quality of vocals and stacking reverb in longer songs really need to be adressed please.
A #1
The issue is that the AI does not remember the original style used. What it does, for each extension, is study the style ~1 min prior to the continuation point to learn/mimic the styling you had (which is why extensions are 1 min long as the other min was reserved for studying).
So if you had a loudness or degradation issue within that minute, it would start becoming emphasized as that is what the AI picked up on and all future extends builds up the issue as long as it exists within the 1 min of continuing.
It also explains why the style can change when you vary the starting point by a few seconds
A #2
If Chorus 2 is within 1 min of Chorus 1, adjust the Extend from time from a second or so and see if that helps.
If Chorus 1 is far away and you want that consistency, it is better to extend from (or even when) Chorus 1 happens and splice in the new Chorus 2 generated part using a DAW like Audacity. (Would be best if there was a small instrumental break or the like so you can have a natural beat to integrate with)
There may be other options too but I cannot think of it at the moment, perhaps trying to search up answers will yield something since this Q is uncommon iirc
oh yes this is so annoying .. but im trying to get used of it. here at 04:50 you can heare the reverb ... i tryed it to intergrate ... tryed to use breakes and interleaves ... at 05:40 i tryed wit "clean voices, recover voices" ... but ... no chance 😄 if the vibe of voices is broken ... there is no possibility to change it back ... but as @fallen thorn wrotes ... try to get more seconds back on extend ... try to find a spot where no reverb is in the sound. this mostly works for me too
https://suno.com/song/ba417e14-f136-4842-ba13-094be8e3e2ff
Trance, ambient and house song. Listen and make your own with Suno.
To avoid this problem, take another song to generate an extend without pressing create at the end. That stops Suno from continue the last song you extend. The Extend Bug have To be solved NOW… It’s crucial. So much credits will be lost….
Nice answer 👍 Wanted to add that the chorus can be more than 1 minute in the past, but you need to be careful how you arrange the composition of each part. Example: if part A has the chorus between second 0 and second 20, part B can have the chorus between second 40 and second 60 of its own duration. Then, when you merge them, there will be 1 minute and 20 seconds between them. Those are just workarounds, I hope in the future the context will be bigger and maybe configurable (for example, let me continue a 5 minute song using as context the first minute).
this happened right after I published a song, it got fixed now (not sure if by itself or after gettings a like), but it was at -1 likes at first
I cannot accept the terms for the lyrics because the button is simply disabled
I find playlists are very buggy, i regularly end up with the wrong songs in playlists
Forgot to tell that you have to press the clear button also after generating the “2nd Extend” which you only do to be able to create a new song and not another Extend which the bug does.
Can you enter lyrics after hitting close on this modal? Will look into this one, thank you
If I close the window and attempt to type again, the same pops up.
Ok thank you, looking into this one
Had same, disabled shield in Brave for a sec and it helped.
hi suno team!This BUG appears again, and it exists all the time. It looks like I can't delete it.lol
hmm, do any of you also have the problem that the cover images randomly disappear? Now that there is the generate button, I can create new ones, but I feel like I do that three times a day?!?
link to a bug I've also experienced
most of them are specific to the mobile ui
I also want to note that the cover image for my album late bloomers just will not stay on no matter what I do, so I have a placeholder
https://open.spotify.com/album/2z3tPbsG8FlIY8N6Dwuptx?si=pbTi6XLYQZup4iuZdb6kIg
presumably there is some sort of basic image detection that sees this as nudity
although I didn't get any message about it, it simply immediately reverts.
what was happening right now if i make a song now there is no intro even have [Intro] and if i extend it the part 1 almost all was creds refunded cause its only 3sec
https://suno.com/song/b2642d9c-2b8d-409d-a5d4-9c37a22960a2
Seriously, no matter what I put, it always sounds like this. Shitty pop you'd hear in a grocery store
https://suno.com/song/fe5bdb04-d402-478f-8227-8291a7e89f08
I avoid the word pop, I still get pop.
https://suno.com/song/ad108c58-d44d-4229-b2dd-4ffe1f47f561
Fucking useless, ask for male vocal, get female vocal
get this occasionally while trying to log in, persists for some time, fixes on its own after an hour or so
i have the same Problem, every song i try to generate, has a completely wrong key in the vocals... it doesnt fit the rest of the track... and its always the same high pitched key like in your song... I cant use the AI like this... nothing helps. And yes the reverb is way to much too
Search doesn't seem complete yet.. If I search on "opera" I get a bunch of hits, but know songs of people who have "opera" in the style of music I don't see in search?
And those songs ARE public, so that should not be the problem.
Let me know if I need to DM you examples. I don't want to post it here so people feel neglected if search is not picking them up 😉
extend not working. It just makes some new random 'noise' and only 12 seconds long
You need to have something in the lyric box, leaving it empty will often result in the shortest song it can make.
The issue is that the AI does not remember the original style used. What it does, for each extension, is study the style ~1 min prior to the continuation point to learn/mimic the styling you had (which is why extensions are 1 min long as the other min was reserved for studying).
So if you had a loudness or degradation issue within that minute, it would start becoming emphasized as that is what the AI picked up on and all future extends builds up the issue as long as it exists within the 1 min of continuing.
It also explains why the style can change when you vary the starting point by a few seconds
I just got an email from support@suno.com that one of my tracks hit 1,000 plays, but NONE of my tracks are public. How is this possible?? I think something's messed up in a database somewhere.
huh, could you link your suno handle?
Welp I know what you guys can do for your "1,000,000" plays award you send to the people that hit it:
or are going to do 
Hands Animation bug
The extend UI is partially broken now, it's constantly resetting the style of music to the style of the extended song.
The 4 minute clips are not working, it even makes them less than 2 minutes
Anything I extend from v3 with v3.5 usually tries to conclude the song altogether. the majority of extensions are coming in under 20 seconds
If you click on these add tag buttons after the character limit in the box is full, they will be added but an error will show up when you press the create button.
yes
I still seem to get the high pitched noise sound get progressively worse, it used to be with each extenteded track but now I thought with V3.5 and the longer tracks that might have been settled, but it seems to persist still even in a single generation, its not always but often enough to be annoying especially its on a song you like.
Profile picture is buggin fr rn https://suno.com/@partvisionmusic
with a link
I have seen a few others pointing out similar stuff with theirs tonight.
Seems like its only with the default - sunoers (not custom images)
https://suno.com/@namestaken
Yeah im on free plan
For me it shows up for maybe half a second (really fast internet) then it breaks
musicalfruittunes
thats not a link 😏 ... thats a link https://suno.com/@musicalfruittunes
Ripping farts for the arts. https://www.youtube.com/@MusicalFruitTunes
WAVs seem to be actually just MP3s at the moment. I'm sure it'll be fixed soon enough, but just a heads up.
Just tested with this:
https://suno.com/song/73882671-9383-447d-9521-cdcf944d8d16
Choosing to download it as WAV gives me the MP3 file but with a .WAV extension.
I can get the true .WAV file by changing the .MP3 URL to:
https://cdn1.suno.ai/73882671-9383-447d-9521-cdcf944d8d16.wav
So I expect somewhere internally they're referencing the wrong source file when providing the download.
suno for some reason keeps generating only 10-20 second songs when I extend, happens with v3 and 3.5 songs
Out of curiosity, did you provide additional lyrics/directions upon the extension, switched on instrumental mode, or left the lyrics field blank eith instrumental switch off when extending?
I tried on instrumental and also gave it some directions still the same result, this doesn't happen when I extend stuff with v3 it seems to be only an issue with 3.5
Also another little interesting thing I found, the past 8 times I extended (with different directions and settings) it was short generations but when I added lyrics it suddenly turned into a full song, it seems unable to extend songs that don't have lyrics, probably because lyrics forces it to be longer idk
thanks for the report, should be fixed now
I thought it was just me and my mixed genre choices, getting a growing distortion and a sort of metallic drone that gets worse, had hoped 3.5 would have solved this, but get a great clean and clear start, but as soon as you think, yes this is the one, the drone and distortion starts, and it becomes harder to listen to part way through, here is one example I've not deleted https://suno.com/song/feed93e6-581c-4c9d-a3b3-3b6ed0a35640 it is noticeable at about a 1minute in, but worse at about 1.55, then as it changes it cleans up a bit, but seems muffled a bit and speeds up in the outro, like its rushing to finish.
Mysterious Progressive psychedelic electronic folk rock, catchy chorus, Clean mysterious male voice. Icelandic accent. song. Listen and make your own with Suno.
I'm noticing the metallic noise still, so I don't think the early access 3.5 is meant to address this - BUT be sure to press the "dislike" on the song because based on #📢┃announcements it looks like they are actively going to use this to try to improve 3.5 for when it gets fully released (whereas normally I don't think this is used for anything - so at the moment it will make a difference)
that metallic noise seems to be too many of the same frequencies accumulating together in the mix. If you edit the track in Audacity, Adobe Audition, or a similar audio editor, you can use a Spectrogram view to see bright spots and then use a Notch filter to remove the sounds in that frequency range. This has worked for me some times, it's not completely reliable.
So far, v3.5 really does not want to continue any of my existing songs. I've picked a handful of different things I've generated over the past few months and it's basically giving me 10-30 second outros in most cases. Just now I tried to extend an old v2 song and it gave me 2x 10 seconds with a couple of notes and then ended. These are all instrumentals, or non-instrumentals with no lyrics in the box but just [tags].
This is happening even if I change the "extend from" time to be an earlier time, even if it's in the middle of some music. Unsure if the lack of lyrics is making it think there's nothing to sing about so it wants to wrap things up?
also helps if you can isolate the stems first and edit each stem to try and remove the noise. https://vocalremover.org/splitter-ai is pretty good
Split music into separated parts with AI-Powered algorithms
Thanks. I've used Audacity to do spectral analysis before but hadn't seen/tried this specific view. I'll try this next time I'm doing some editing. I mainly wanted to highlight the use of the "dislike" button though as it seems this is going to be actively used to help address the problem at source (i.e. in the model.)
very encouraging that they are actively working on the issue, to me it is Suno's biggest issue
Profile button disappears in landscape mode on samsung galaxy zfold 3 when phone is open for larger screen.
https://suno.com/song/5d51ec4a-3a17-470e-bd7b-828c8bb15c01 this version of my song, something starts going more amiss at about 2 minutes in, and slowly gets worse to the point it disappears about 3.18 before the lyrics have finished, just fades away. Have thumbs down it.
Ambient progressive folk rock, Clean mysterious male voice. Icelandic accent. song. Listen and make your own with Suno.
Not all submenu buttons show up on screen a large majority of the time, especially the further past center of screen you are. Sometimes not showing any options at all.
In addition to v3.5 trying to end extensions of v2/v3 generations early, it wants to add a piano into almost everything when extending them.
I'm trying to renew my subscription (monthly) but whenever I try to send the bill information, I get this pop-up. My ad-blocker is turned off, and I tried on both Chrome and Firefox, with the same result. Anyone's got an idea ?
not sure if it counts as bug, or needs to be added but 3.5 custom till generates lyrics based on 3.0. So only 2 verses and 1 chorus, vs basic that makes 4 verses and 2 chorus, it would be nice if we could generate longer lyrics using lyric generation is custom
hmm. having some instances where Suno is adding vocal snippets that voice the 'Style of Music' prompt terms?
very beginning of this for example:
https://suno.com/song/88a7646e-9ddc-4284-9eed-3ac1d220159b
neurofunk, uk dubstep, growl bass, minimal, bukez, sublow, oscillator, 140, 2-step, filthy, dark, taiko, downlink song. Listen and make your own with Suno.
version 3.5, by the way ❤️
yeah doing african chanting I was getting some chanting words before he song started
Which side? You can see which songs were selected
What I’m talking about is where can I see the works selected for the competition?
Hey, I made a few tracks and I turned on the "Instrumental" setting, but a couple of them still generated lyrics.
I bug reported them for "Other" because there didn't seem to be an appropriate tag for that scenario.
Also it may be possible that the error was with the website frontend rather than the model or something.
Just wanted to provide additional context.
songs generated by others are randomly appearing here, is it a glitch?
yes. these should be songs from the explore page. sometimes there is some kind of caching bug. if you refresh the page they are gone
Im getting a lot of 0:15 seccond to 0:45 on the beta. It seems like the faster and more complex it just breaks - i cant even extend my songs and im not getting refunds for any of these
Okay it keeps resetting my time stamp when I’m trying to continue and it’s super annoying and I’ve wasted about 20 generations so far
yeah its not working good at all, i cant extend anything it just gives me 15 seccond clips
its not following any of my prompts, and they work fine in 30
Not sure if its a bug or not when i can make 2 songs from 10 credits pressing the create button while the first one loads
WAV download much appreciated! But it seems like anything longer than 5:00 winds up with an error message. Is there a time limit on wav downloads? Or is this a bug?
I feel like the tempo is still trying to get through the song in 2 minutes even though it’s now 4, it keeps rushing and then quitting
I have kind of noticed a 2 minute bump or shape as well, though not sure
No longer recognizes the style electro swing, same song works perfectly every time on 3.0
I noticed this mistake a long time ago and it's super annoying at the end of the song it somehow drops a lot, please
Electro swing works when entered as electro-swing. However the four minute length is simply not working at all. https://suno.com/song/90a445e8-1d5c-46ac-840f-0296c51ffc53
Electro-swing song. Listen and make your own with Suno.
Has anyone successfully gotten a four minute generation?
This is the other Gen from that prompt. This sounds like hot garbage, this song works perfectly every time in 3 https://suno.com/song/fc4f6400-c125-421e-b801-adba3bc080d9
Electro-swing song. Listen and make your own with Suno.
Look guys, I just ran this prompt several times in both and my non-musical husband from across the room can tell 3.0 from 3.5 because 3.5 sounds terrible and 3.0 sounds great. Y’all need an intervention
Regarding the reported problem of getting an mp3 file when choosing .wav format: I just downloaded an old generation I had from the v3alpha days, and got a valid .wav file that contains more information than a corresponding mp3. It has uncompressed audio, and there's audio content all the way up to 20000 Hz. 👍
So.... bug appears fixed.
I'm also puzzled how it was able to regenerate this v3alpha song. 🤔
it still sounds as fantastic as the original v3alpha!
so maybe the v3a engine is still in there somewhere
I know it says it's "generating" when you choose the WAV download option, but it's pretty quick - I don't think it's actually using v2/v3/v3 alpha/v3.5 at that point. I expect it probably already has the raw audio data stored somewhere and is just creating a WAV file from that (and then making it publicly-available.)
it took enough time that I think it was actually doing more than just recalling an existing file.
so no, i don't think it had the raw data stored tbh.
also, that would take up a lot of space
but yeah... hard to say... suno is completely undocumented
No bagpipes, no gospel, no 60s garage band, and my husband keeps saying “wow you can still use the old one right?” Whenever he hears a 3.5 version of a song of mine he recognizes. Also when I ask for an instrumental using the instrumental setting I don’t expect it to be 47 seconds long and nothing but singing.
Just got a six second long instrumental using the instrumental setting https://suno.com/song/dc284a76-7599-4ada-87e5-0277846168c9 at least it sounds kind of like a bagpipe. A bad synth bagpipe.
I have not tested out everything.. but Jonathan created bagpipes: https://suno.com/song/7ae41ac8-67fe-461f-93fd-2ede56db6e43
Scottish-bagpipes, Scottish-bagpipes, Scottish-bagpipes, song. Listen and make your own with Suno.
There's only so many times I can keep reporting tracks (which I pay for) as "low audio quality" and "doesn't follow style prompt" before I say "f*** this". V3 or V3.5. it's all the same problems. This is still very much a WIP.
Time to go and touch grass...
@neon ivy garage band https://suno.com/song/90476de2-fdb0-4bf1-b209-bb3eda586a51
gospel https://suno.com/song/a5b9537e-cb27-475c-9832-4d27b6dd1094 (this was the first roll and I don't like the quality but it shows it can be done 😉 )
60s garage band, sixties-garage-band, sixties garage band song. Listen and make your own with Suno.
black-gospel, black gospel song. Listen and make your own with Suno.
This is very frustrating, especially when it gives you a piece of music that would sound perfect if it weren't for the audio quality degrading.
Specifically for poor audio quality: When testing v3.5, press the thumbs-down/dislike on anything that has poor audio quality - this will influence how the final v3.5 performs when it is released.
It's been very genre specific. Been with previous models too. I have songs without asking for an intro, and getting it, and some I am asking for it, and not getting it.. Here an example.. not asking for intro, still getting it: https://suno.com/song/746d4b56-1314-47d6-a96b-23cf18ac4123
minimalistic, clear voice,Symphonic Metal, Alternative Metal, Symphonic Rock, Gothic Metal, Hard Rock, Alternative Rock song. Listen and make your own with Suno.
I personally use the instrumental option, if I am not getting an intro.
May I ask for your "style of music"? Because instrumental works with my prompts. I love to understand the difference and see what works and what not.
Alright.. Let me do some testing myself.. Will let you know the outcome 🙂
I think I already found the magic metatag for your.. try [instrumental hook]
https://suno.com/song/b9f52559-20f2-4076-8d96-52090c0ffe01
https://suno.com/song/def820a9-836d-4c0d-9a01-4939a4489fe3
(still rendering but both with intro)
Yeah I like them too.. You've got great taste of music.. A new genre added to my list 🙂
I have to do some RL chores so will be afk, but will check in later so see what could do the trick 😉
what i've noticed on 3.5 that the longer the song goes on, the more distorted the music and vocals become, not sure if this a common experience but in my tests it degraded much faster than in the previous maximum 2 minutes of v3, even before the 2 minutes prior limit
especially vocals and higher pitched music starts to become more and more echo-y and scrambled at like 1:30 or earlier which isn't ideal
I'm encountering this quite a bit as well. Be sure to downvote clips with poor audio. It won't make a noticeable difference right now but they're using this as feedback.
time to downvote pretty much everything then lol
because I have yet to create a track that doesn't get completely scrambled towards the end
Bug on mobile. I cant see the new download menu with wav
Still only getting tin-can sounding rubbish. It used to have some actual interesting sound. Guess that's gone.
It doesn't let me accept. I'm on a Pixel 6 phone.
I'm going to downvote this one due to audio quality but I felt it was worth specifically sharing this v3.5 generation here because the quality after 2:00 is atrocious to the point that it becomes over a minute of just noise.
Additionally, as others have mentioned, instructions to add an instrumental intro are completely ignored. So, I have taken to wasting credits on starting a song with instrumental only and then extending with lyrics. Unfortunate, because the first generation is the longest in all versions, including 3.5, but starting the instrumental intro there means I toss at least half of that generation to extend from an appropriate point. And now, in 3.5, a problem that showed up in 3 is worse: instrumental generations have lyrics and other random vocalizations. 😵💫 Please please please fix "instrumental" prompts.
Playing with instrumentals and I’m getting a lot of what someone called “shedcore,” where it’s just the sound of someone beating a metal shed with a baseball bat. This is a good example, the piano at the beginning is not what I asked for and there’s no reason for it to exist, and then it devolves into shedcore. https://suno.com/song/71b05892-d670-4a11-bdf7-6ca6e1f7a6af
dramatic space-themed military soundtrack, Orff, Holst song. Listen and make your own with Suno.
not a solution just a related idea. Testing live music in 3.5, lyrics has dialog introducing the song which has a decent change of generating something like a song intro as a side effect. Playlist https://suno.com/playlist/ee70007f-261e-4b7e-be95-ad6c14386ed8 Maybe you can flesh out the 'instrumental' intro with a few stock phrase, even ahhs and ohhs, and get something similar.
What are all those symbols??? I like your playlist.
does anyone know how to get clean sounding synths/303s? they allways sound like they've been bounced 20 times on cassette. Also i notice a lot of nasty high frequency audio problems in my dance music generations
i keep reporting them as bad audio quality bugs... i hope it can be fixed
It's sort of joke song where the lyrics use a 'table flip' cartoon you can draw in text like this: (╯°□°)╯︵ ┻━┻ it's a face with eyes, flipping a table.
But when Suno tries to sing it, the singer just screams or says random words, and I get a lot of joy out of how realistic the songs sound but how absurd the actual content in the song is.
https://suno.com/song/4baa3c6e-3930-4e36-bbda-c7bcf66f7eef
Is it normal to hear a discord notification in the song at try between 21-24 seconds?
Rock | Beats: Heavy, driving | Tempo: Fast, 140 BPM | Mood: Energetic, aggressive | Vocals: Bold, intense song. Listen and make your own with Suno.
I'm not hearing it
is it the same with you that the instruments are always overtuned from the middle of the song onwards? v3.5
overtuned?
Example of a newly-created v3.5 instrumental that isn't even 1 minute long: https://suno.com/song/50ae1193-be1b-4ead-bd22-b1ba36811419
Most instrumentals are way too short. Some of them sound quite nice though so I don't want to share them here 🙂
Yeah with instrumentals ive found you have to make a proper cue sheet to get them to four minutes
4/5 lines per minute is usually enough
And they often start losing quality a minute or two in
like the one with lyrics. hope this is solved in v4
I am back behind my PC.. and I am going to do some tests.. but 1 thing came to mind, that I was not sure you knew. You can use the instrumental toggle during a song. This helps sometimes to create instrumental and solos if the metatags are not doing their job. I personally use that a lot. Did you already know that?
v3.5 - Instrumentals that manage to generate > 1 minute do seem to start getting tinny/metallic around that point.
getting <30 secs on Bagpipes, but still takes the Credits WHY?!!!!
enjoy the credits
Silence is how suno makes their money 😂
We get more bang for our buck now, but we cannot win them all 😄
Because it's over 8 seconds? I believe around that is the threshold
that is to short of a threshold, it needs to be around 15-25?
i think everything that is beta and the community provides feedback should only pay 50% if at all
That more a suggestion I think
I am not sure how long you've been here. But in the beginning you only got 100 credits for 10 dollar... So that is more than 50% better what we get now
I mean.. I would LOVE to have everything for free. But this company is very generous imho. And you don't HAVE to test. It's fully up to you. 🙂
maybe, but i just spent another $8 on Throwing the song away:)
not agood
Sorry you feel that way. But keep in mind, you don't have to use it. And they are pretty upfront about the quirks. If you send an email to them, I am sure they will refund you some credits. 🙂
If they reduce the price of the pro and premier then I'll consider buying it
i generated over 40 songs yesterday and can't use almost any of them because the instruments in the background get too shrill after 1 minute. i don't want a refund, but i also don't want someone to tell me "don't use it then".
I understand. In that case, thank you for your feedback. I know it is not perfect atm. I wish it was too. But devs are working hard on V4 which will not take too long as I heard from them. So please stay tuned!
Another especially bad v3.5 instrumental. It generated 4 minutes but very quickly just turns into noise. https://suno.com/song/ba7ce28a-67fd-480c-a576-a81e6fc15851
Gotcha.. I am on it!
Model version box needs to be made wider on mobile
I think I found the solution for you. Continue where you want, empty your lyrics and only put in [instrumental hook]. It gave me on the first go 2 perfect instrumentals. Then just continue again where you want the solo to end and add your lyrics again.
Here the full song I made for your test. The instrumental solo starts around 1 minute: https://suno.com/song/4173f16b-3b0e-4535-a21b-b6bdfb6ef0ba
nu-metal, hardcore-nu-metal song. Listen and make your own with Suno.
┬─┬ノ( º _ ºノ)
More reverb hell stacks on extensions and (metal) guitars, riffs etc.? I have my weekly song for my channel for this and next week already done thank god, gonna wait for a bugfix for now 😉
/Chirp
I'd like to say that since the 3.5 update I've yet to get a 4 minute initial generation. Most are around 1 minute and I'm even get as low as 9 second initial generations
Did it not run through all your lyrics or were you just doing an instrumental?
I've consistently have 3:30–4:00 songs today on 3.5a. Have you given it enough lyrics to sing?
For instrumentals you need to write a cue sheet in the lyrics to get the full four minutes
How do you do this? Do you have an example?
something like this:
[Ominous Bassoon]
[Gentle Piano Arpeggio]
[Dissonant Strings Enter]
[Frenetic Orchestral Accents]
[Piano Melody Returns]
[Swirling Piano Flourishes]
[Timpani Rumbles]
[Brass Fanfare Builds]
[Chaotic Orchestral Frenzy]
[Serene Piano Outro]
[Minimalist Piano Intro]
[Flute Echoes Melody]
[Tension Builds]
[Crashing Cymbals]
[Tranquil Piano Interlude]
[Brass Swells]
[Urgent String Ostinato]
[Clarinet Shrieks]
[Thundering Timpani]
[Delicate Piano Coda]
I start by picking an existing song as an inspiration and ask ChatGPT to come up with a cue sheet for the song (it's usually way off, but that's how i get started)
Right. Even if the AI doesn't understand well, adding something like this usually at least adds length and can also make it less generic, if you were having trouble with that.
I gave it a try with some obscure/weird tags which, as I expected, gave me a ~2 minute piano-driven piece of music whose quality gradually got worse. And I've just tried it with some more serious tags (verse/chorus with some direction underneath) which were mostly ignored and again I got a ~2 minute piece of music whose quality gradually got worse.
To clarify, I don't think that type of prompting helps with audio quality problems. It's more the problem of the composition being repetive, that kind of quality.
I was mainly trying to get a longer instrumental.
BTW, in that specific prompt, probably remove the outro in the middle, just in case that trips it up
v3.5 instrumental: https://suno.com/song/c4a36321-fe61-469f-8bbc-782cc9c26a55
Seemed to ignored most of the prompts, produced a 1:17 long song. Unsure whether I should consider this poor audio quality - it's mono throughout and I think it starts to ring slightly toward the end but I'm not sure.
Unable to press the "I accept" button on mobile, locking out of custom songs.
I haven't tried this V3.5 but in V3 I sometimes just fill the lyrics with whatever and seems nice:
https://suno.com/song/28c314f4-eeab-4381-98e5-fb4f4adb1e56
close, fugue baroque, technical weird-trumpet, *, martial,indie, binaural-synth, V3, V3",," acoustic-casio, spare song. Listen and make your own with Suno.
have you tried to empty the cache or reload the page?
Interesting. First reload no, close browser and open again yes. Likely not reproducible. It works now though.
Is there way to contact support? Tracks over 10 minutes are not able to be downloaded as wav for some reason.
Filling the lyrics box with dots and dashes does seem to force it to go the full length. It is however also hallucinating vocals when I do this.
i had no problems downloading this song ...
Yeah, I forget not everyone loves that. Repeating "instrumental" in the style prompt a few more times may help, as well as [instrumental] placed at intervals in the lyrics. Maybe roughly spaced like verse tags.
I know 3.5 is early access, so this is probably already being worked on, but I'm getting some very bad fade-out of vocals over the course of a long generation. E.g., https://suno.com/song/6f2ce02d-43b0-4eb1-aec8-deb830401b9c compare the vocals at the end to the beginning, they're barely even audible, and sound...I'm not well-acquainted with the terminology of this, but they sound very fuzzy, hazy, as well.
electronic, violin, violin virtuoso, angelic female vocals, trance song. Listen and make your own with Suno.
More than welcome.. Have fun creating new songs!
"spaghetti western" seems to have disappeared in 3.5, I tried multiple phrasings, tried pushing the relevant instruments into it, and no dice... it focuses too much on 'western' and ends up with general country ballads
Tried "spaghetti-western"?
yeah, and spaghettiwestern and spaghetti, no dice
it still understands the western in every part, though... but as 'country' now
or rather, western-country
Found thiso on my first ever song. WAT HAPPENED??????? DOES EVERYONE HATE ME????
It seems impossible to extend an older (non-v3.5) song with a v3.5 instrumental, even when padding the lyrics box out with symbols it immediately wants to end the song (with a long tail lasting 10-20 seconds)
so far V3.5 is a really bad experience for me. the biggest problem is suno tries to rush everything. the singing is way too fast, I put 4x verses + 4x chorus + instrumental prompts and it finish all in about 2:55 - 3:05, where the instrumental parts are way shorter then in V3, the sound is boom boom boom fast music fast singing short instrumentals (if at all) boom finish. also found genres aren't read well at all. trying "metal ballad" which currently sound well in V3, but in V3.5 it's just speed metal. end results so far are extremely underwhelming
The syntax on styles seems broken, it doesn’t recognize multi-word phrases like spaghetti western or electro swing
So, after a day using it, I have to say 3.5 is pretty broken right now and I hope 3.0 remains available. Throughout the day I kept having the experience of saying, “Oh that one turned out good, maybe I’ve got the hang of 3.5!” And then noticing I accidentally reset it to 3.0 so it was much better. It doesn’t do many styles it used to do at all, other styles that used to be good are bland and generic now, and everything is coated with noise after 30 seconds. But kudos for adding didgeridoo.
Seriously it’s awful it needs fixed.
Please do something about the out of control watermark. I was hoping that in version v3.5 it would be corrected, but unfortunately it is still the same, i.e. after some time the watermark becomes so loud that it drowns out almost the entire song 😦
Cover image for a song doesn't save when I change it.
I refresh the page and it goes back to what it was before.
Hey - I think suno has some troubles generating dark and mysterious ambients and songs. Everything is "happier" than before (?). I'm using Suno to make music for my game which is placed in the mysterious and dark world and previously Suno was fantastic in generating such a fitted music. What happened?
Also no matter which model I use (v2, v3, v3.5) - results are worse than before
With v3.5 - I'm attempting to make a longer instrumental by stuffing tags in the lyrics box.
Part 1 was 1:19 long, and starts winding down from 0:50. Final 10 seconds are just quiet noise. https://suno.com/song/ea573da6-c6bf-48ac-8e67-2a1126aacde7 - some metallic ringing around 0:40.
I extended from 00:30 and left the same tags in the lyrics box:
Part 2 (gen 1) is 22 seconds long and ends at 15 seconds, quality is fairly poor https://suno.com/song/751f00d8-70a3-42b5-a470-2f566b34c1ba
Part 2 (gen 2) is 57 seconds long and tries to end at 15 seconds with a load of noise. At 30 seconds it starts playing some music mixed in with the noise, then gives up again at 45 seconds. https://suno.com/song/554229ea-124f-4416-ae11-312a86c92a8c
V3:
I don't like the long pause in the extension (https://suno.com/song/bc9406b1-f51e-459c-bb04-3501e18f8345)
Alternative Rock song. Listen and make your own with Suno.
Otherwise it's good
Tested extending an old v3 alpha song with v3.5, with just the instrument toggle "on" so nothing in the lyrics box.
Original v3 alpha song: https://suno.com/song/478787c5-f5e9-4e12-8465-0a9c1bda93cd
Extending from end of clip:
1 - Song ends quickly (credits refunded) - https://suno.com/song/838013de-81fa-4399-b5b9-86bf31651130
2 - Section of music ends, then sings the lyrics from the start of the original song with a quiet piano accompaniment, song ends after 26 seconds - https://suno.com/song/afd2db4e-6750-4861-b3f4-4ec7b3d601d8
Extending from 0:59 which is in the middle of music:
1 - Sings lyrics from part 1 at 0:15 and winds down the song - https://suno.com/song/89c3cc18-62ab-4a08-8aa8-a297c26d5bab
2 - Weird vocals in background from 0:17 which seems to be partly repeating the lyrics from part 1 whilst the quality gets worse. Last 10 seconds or so is just basically noise. - https://suno.com/song/fd7219f1-3b47-4621-923c-92e2bc57534c
Basically with the instrumental toggle "on" when extending an older song with v3.5, it seems to want to try to end the song as soon as it's reached the end of the section that the "extend from" time is in.
Starts to degrade at around 0:30. Poor quality for sure.
Trying to make tracks in Instrumental, v3.5, and the length of 1st gens varies so much:
2:19, 0:34, 2:22, 0:03, 1:04, 0:49, and that's just a small selection. Hopeless trying to extend from those short gens. Not getting many credits refunded for the very short ones.
I've had a few gens in v3.5 where I've prompted for Orchestra and got piano!
"Piano" has always been Suno's fallback when it doesn't know what to do, so it's a bit concerning that this seems more common in v3.5
I've tried that too, without much success. I filled the lyrics form with [Instrumental] tags in the hope I'd get a 4 minutes instrumental, but it didn't work. I got about 2 minutes, I think.
I have gotten a 4 minute generation but it took a while before I finally got it, it doesn't give you any control over how long you really want it
Is renaming meant to be locked behind paid subscribers now?
Edit details
Is for all
which section? I am blond after all lol
PC or phone?
PC, havent tried on phone
Hmmm.. maybe I am wrong.. Let me check this
That's really weird. I can change it on a free account. And a dev is looking into it now.
maybe i was too finicky about my titles so the servers auto locked me out lol
This is what you "should" see
OH WORD, i AM blond lol --- you can only do it from "Library" now - and it looks different (like you show)
thx
No officially you should also be able on the create page too.. if you don't see it, it's still weird lol @spring heart
So still good you said it
Yup. I have a song that ran for 4 minutes, v3.5, despite that the lyrics took up less than half that time. It did an interlude and repeated but by the end had a weird humming noise. Extended from 4 minutes with new lyrics and got credit refunded on one generation and 9 seconds of music on the second. It did not sing the additional lyrics.
Extended from slightly less than 2 minutes and got the song to extend as intended. But both extensions were crapfests by the end of the first minute of extension. Same fuzzy static hum in the background.
when i open the 3 dot menu on the home page the lyrics are ON TOP of the menu... looks funny 😄
Please.clear up the voices and make them better. I do use the competition too ans the.voices are better. I do believe suno u can take over of you fix the voices.
are you saying that is a bug?
playing with v3.5 and a lot of my extends are 15-25 seconds and not very useful, burning through a lot of credits 😔 Have tried extending from timestamps on both v3 and v3.5
did you try continue from middle of it?
yeah I've been trying Extend with MM:SS timestamps, I usually try to find the timestamp where it could transition into a new flow but it seems to try to end the song as fast as it can, even if it's just extending a minute of part 1
Same for me, timestamp ends quickly. If you try to create 3 and 3.5, a short one is 10 seconds and a long one is 2 and a half minutes. If you use expansion, it only expands by 10 or 20 seconds.
3.5 is unusable garbage for any song with any actual effort being put into it.
-
Eventually it randomly adds a high pitch tone that overshadows the vocals entirely, for literally no reason. The sound is not present at all before it shows up.
-
The volume of the vocals is ridiculously low. To the point where i could do a heavy metal or sub genre song, put various meta tags for screaming, shouting, yelling, ect, and the vocals are still significantly lower than the music in the background itself.
-
The actual vocals themselves are largely poor. Randomly switching vocalists or vocal styles unprompted, being more robotic, or blending in with instruments entirely.
-
The longer memory means more generic boring music with little to no control over how the song progresses or changes.
-
Style prompts suck now
Something like this is literally impossible in 3.5, and in fact some of the above issues leak into 3.0 as well.
I filled out custom lyrics and style and hit create and it generated the random non-custom thing. I want my credits back. 😤
Refreshing page fixes this
Sure. But meanwhile I wasted credits. Nothing about my setup suggested it was going to do whatever it wanted to do, instead of what I asked.
Hi -- I heard your feedback. If you find the noise /quality issue, downvote on the clip is the best you can do to help.
Style prompts, at least for metal, are very poor. Most of my power metal songs end up as extremely bland pop rock sounding vocals no matter what prompts are there. The creative and interesting gens (at least when lyrics come in) seem to be seriously dumbed down with 3.5. Everything feels like it was tuned to be extremely "safe" sounding.
Hoi folks. Just tried out 3.5. First feedback: the stacking reverb/delay issue got better, but is still present. Specifically when I extend from section breaks, where a previously triggered FX might still be in decay, the stacking comes in effect again and gets notably bad after 3 or 4 more extensions. Recommendation: generally shorten the decay of reverb and delay, UNLESS specifically prompted
Have you tried putting specific vocal instructions into brackets of the beginning of the lyrics?
Yes, I tried all the tips people have suggested for 3.5. I'm done bothering with it for now. If it isn't boring and bland, it just turns into some audio that sounds like it is coming from the neighbors house within 2 minutes.
also like this?
[Verse, raspy male voice, g-major key]
blablabla,
Has worked several times for me, but propably needs some runs. In case you wanna try
v3.5 Vocals are still very extremely noisy. Hope this will be fixed by v4. If it can at least match udios, it be sort of ok.
Iteration 5, we're getting closer
https://suno.com/song/9a619850-cabf-4184-a73c-878cd1c096bd
Iteration 6, well... too much credits used by now for just an "experiment"
https://suno.com/song/80a359f6-25d8-4c5c-9653-c39fd47cec56
Sure, you will not get George Fisher out of it, but I guess a few more runs could potentially yield usable results
dark metal song about to much reverb sucks song. Listen and make your own with Suno.
dark metal song about to much reverb sucks song. Listen and make your own with Suno.
I'm confused, are you trying to show me how it is ignoring the prompts and tags?
With the second one 
nono, yeah, should have posted the earlier runs as well. It is like you said, everything pretty much starts out as teeny-pop garage band. But it gets closer to what you want over several runs. Unfortunately I already deleted the previous iterations
Yeah I have gotten ones that I like, but it is a lot less common.
Anyway I don't think this is the spot for discussion
It seems that ground is being lost to Udio here. Despite the many recent innovations, the songs are now faded out after a section, at least in the free account, or a blank space is created, so that without [Outro] you think there are 3 songs. In addition, the [Outro] is faded out, but strange noises are inserted after an empty section, which can even happen twice in a row. On top of that, the passages suddenly have a significantly poorer quality and are quieter.
If you already don't have the quality in the free account, you'll think twice about whether this shortcoming is probably also present in the paid option, even though it's version 3.5.
3.5 is so much better than 3 if it didn't include the clatter in every attempt. As it is, it is currently unusable. At least for music with a band. It might be more suitable for electronic music.
The sound is too low, hard to listening😂
https://suno.com/song/4cdc9cc8-5a8e-4951-8172-c39b683e0e5a
cinematic classical gothic rock song. Listen and make your own with Suno.
Somehow Suno really likes the 'untold'/'unfold' rhyme, has anyone else noticed this?
Just wanted an instrumental. Now suno wrote a love Song 🤷🏼♂️
No lyrics included
Tropical-House, Deep-House, 130bpm song. Listen and make your own with Suno.
https://suno.com/song/2b9392fe-0cfd-45ad-8f8c-ef7f46449112
Next one 🤷🏼♂️
Tropical-House, Deep-House, 130bpm song. Listen and make your own with Suno.
plz flag with dislike
Done and flagged as bug „Not following instructions“
I'm trying in vain to create clean, cold electropop like De/Vision or Mesh, and it's genre, vocal and any other attempts at steering the prompt into my admittedly narrow genre vibe is frustrating at best. Not getting anything decent. Vocals are always too high regardless of putting in baritone, vocals sit too quiet in the mix, and the sound quality just isn't there. I'm fed up of the reverb washy half spoken close mic vocal style it throws in everything. If I wanted Billie Eillish I'd ask for it! I think it gets drums pretty well, but beyond that this v3.5 is awful IMO. I guess I need to wait a year before it can generate anything useful for me. Cancel sub time I think.
Any one else got problems with extensions? Getting 19 und 20 seconds. (I'm still on the free plan).
This usually happens if you're extending from a point where Suno has already decided to end the song. Try moving the "extend from" time to be a few seconds earlier (e.g. just before the end of the previous passage of music) to encourage it to continue rather than end.
The version was already cut off in the middle of lyrics. I went back to a point before to get the whole verse. 🙂 It's the second extension. The other ones got a minute (a few days ago).
Tested a lot yesterday and today... Here is a report from me 😄
3.5
- Tends to output depending on lyrics, not for the style.
- Does not listen to outro or ends, and mostly ends when it wants to.
- Songs come out more "pop" then the style defined.
- Randomness is not good, 3 was better at this. Even when lyrics are altered, song has the same melodic response for the same key.
- 4 minutes is great, but after 2, distortion starts to appear.
- More lyrics = more speed in the song, still tries to finish in 2 minutes.
- Chorus is sang faster then the verse.
- From time to time, half of the song moved to a different genre.
- Heavy piano usage.
- Even though there is a higher vocal and instrument quality, more instruments in the song = bad quality output.
- Powerful vocals suffer in vowels.
- Lyrics are well placed, but lines are merged. Version 3 when used commas, dashes and dots understood what we wanted better.
Currently starting with 3 and extending with 3.5 gives better results.
On 3.5, I asked several times to get a "women only" or "women group" but I'm still getting only male singers/rappers. I had to say both in the style or add [female 1], [female 2] to the lyrics get the results I wanted.
female vocals usually works for me
Yeah. I never had problem with that in 3.0. I noticed in several 3.5 generations.
In 3.5 vocals tend to leave "ssss" sounds if used like this:
[Intro]
Aaaaa aaaa
Aaaaa aaaa
Why is this here?
I agree on some of the above. Been finding some really bad audio quality recently, but in both v3 and v3.5 alike, get some tinny sounding echoing going on and noticed that in cases the first part of the lyrics is being cut off. I have flagged multiple of these in the hope it helps train up the model and improve. Really hoping they solve the tin sounding audio, as it makes for some bad experiences with the outputs. I go like the idea of outputting the whole song though, just needs some work in sections, as mentioned above, once you get in to the song a bit, normally around a chorus hype, the audio goes to trash for the rest of the song.
Can't say I have had this issue once, I just stated male vocals or female vocals and it goes with the right type.
3.5 never follows the style of music I want unless I continue a 3.0 song.
Played about with this and done some interesting types on vocals, have creepy female vocal which worked well, black female vocal also works well and just for a bit of fun alien female vocal and all came out pretty good.
It definitely seems like 3.5 is deciding overall style based on the actual lyrics and theme of them, overriding a lot of the style prompt, which really doesn't make sense.
The machines are rebelling!
Agreed. Maybe 3.5 should take less credits for subscribers. Maybe there are tags we’re missing lol. Or sky net doesnt wanna sing for us.
😂😂
I put reggaeton or bachata and it clearly gives me rnb or trap. So I tried dembow. Still trap. I did try Latin hip hop and boom I got a reggaeton style beat.
Would be cool to change the tags without using credits. It’s very fun to experiment with.
hey all I've found one issue that Suno struggles with. It can't create songs with the singer being the gender you specify. For example I'm creating a love song and no matter how many times I told it I want it sung by a male singer, I usually get a female one
Did you try to add in the "style of music" things like female vocals?
ah yeah it seems to work better if you do that. This is really an issue if you dont have custom mode enabled
Ah yes I personally only work in custom..
mouse scroll wheel doesnt work in the edit profile dialog (chrome / PC)
Artifacts at the End of the track: https://suno.com/song/ba60c581-f6fb-43b5-8669-97c73141b058
Alternative Rock song. Listen and make your own with Suno.
This is an extension of some already existing song
!SCAM! DO NOT LOG IN WITH YOUR STEAM ACCOUNT
What do you mean? These artifacts have nothing to do with the original song, they do not belong there
they sound military
Interesting...!
I think they sound more like early swing music
Sounds like some church gospel going on imo
The excessive Reverb issue on V3 has been fixed in V3.5 but the same issue in V3 of the vocal getting lost in the mix on a Deep House style is the same in V3.5. Prior to a bass drop when all is calm, the vocal sounds like a good level and a front position in the mix. After the drop, the vocal stays the same and can't hear it and gets lost.
I dumped a ton of tokens on 3.5 and uhhh i got a lot of 15 seccond 20 seccond tracks that havent got refunded. Do you even refund stuff spent on the beta and is there like an option for premium users to get their stuff reviewed or something? or do they all get review eventually?
Weird...... it's suddenly telling me I am spelling everything incorrectly! Trying to correct me with another language that, Google Lens says, is in Danish!?? It was working ok, then suddenly changed and it's working ok in discord, weird why only Suno is now being flagged, I have changed no settings.
[BUG WHEN JOINING EXTENSIONS INTO ONE SONG - LONG SONGS]
When you click on the "Get Whole Song" button, the parts are all joined together, however, at the separation points there are inconvenient noises that "dirty" the music and make it uncomfortable to listen to.
This every minute becomes very uncomfortable for the listener.
The noises caused are precisely at the connection points between the extended parts: 16:00, 17:00, 18:00, 19:00...
Apparently these noises only happen in longer songs (over 15 minutes or a little longer).
Please fix this bug so long songs flow correctly without unpleasant noises. 🙏
If you need more details and samples, call me privately, it will be a pleasure to collaborate.
Thank you and congratulations on the excellent work
I don't know what this is but I'm pretty sure Suno doesn't do any spell checking (...or is this a new feature?)
Browsers do spell checking
No, it's a language tool I have on my PC, but for some reason it is trying to correct my English to Danish, but only in Suno, no other webpage is affected. it's just strange why it suddenly thinks I should be typing in Danish only in Suno!
If you're browsing Suno songs, maybe it's getting confused by the mix of languages
I'm extending some purely instrumental music, and it's the 'style of music' box. The Title box is not affected.
This is probably some setting in LanguageTool you can change, like a default language or setting per website. No idea why it's getting confused but it's more likely an issue with that browser add-on than something Suno need to fix.
Probably isn't anything to do with Suno, but just really weird that, as I've gone to further extend the music and that it is only where you type the 'style of music' in, is affected, As where you type the title in, it's not flagging spelling errors in the English Title I've entered in.
If any team members are interested in a House music issue that's been in V3 and worse in V3.5, the vocal get's lost in the mix and some high annoying sound gets louder and louder. Waisted heaps of tokens trying to get past this but no. Will have to ditch what could be a good piece. Can hear it here.
Mine just started doing it as well.
Hi, is this normal? I'm new here.
You can make music via https://suno.com, there is no discord bot anymore for that.
Ahhh many thanks. I just watched a YouTube video, it's probably out of date 😅
Correct.. For more information see https://suno.wiki and https://youtu.be/lj8utBd_IUM
Hello, I'm 3Daizy, a moderator and guide at Suno. Welcome to my channel, where I share my journey and experiences with AI-powered music creation. Whether you're a budding musician curious about AI or an experienced artist looking to innovate, you'll find useful insights and practical advice right here.
Join me as we navigate the fascinating int...
v3.5 vocal quality drops a lot as the music progress. A 3 minute song, #1 is from the start, #3 is from the end.
The vocals suffer a lot.
plz downvote the original clips
@tame latch @delicate iron 3.5 major bugs vocals do not clearly say what you want they mispronounce words 4 minute generations have god awful static noise at the end and also ai does not respect the style you want also the sounds are worse than 3.0 also the music in general is not as quality it is video gamey also you cannot tell ai you want electric guitar solo only and specify flying V or Stratocaster instrument it always even in instrumental mode makes drums also no matter what and the generations in 3.5 take way too long also compared to v3
You have to stop reposting this, including pinging the Devs directly. It's Sunday and they are entitled to a weekend too. @wary flare
For repeatedly reposting including directly pinging devs, you will be timed out for 24 hours.
Hi guys there’s a bug where after you download a generation as .wav it will just continue to download wavs of future generations automatically. (Using Mozilla - haven’t tested if it happens on other browsers).
say more?
huh
that's weird..
I just think I fixed a bug of wav download of v2 files
btw -- no one told me that
I'd love bugs channel to have more bugs rather than complains about audio quality...X.x
Has anyone had any luck downloading a WAV longer than five minutes? not working for me… 🤨
send me a clip that you want to download it from?
Maybe need a rank or subscriber type that are willing to look for bugs?
the link to the clip would be good
Will do, give me a min.
k I will wait for this and then go to bed..
Downtempo, Arabic percussion, dubstep, hypnotic song. Listen and make your own with Suno.
This one is quite long, but I have tried clips down to six minutes and that doesn’t work either.
let me try
5 min. seems to be max.
Hmm… will look.
Nope. “Download failed” message again.
I’ve tried many different clips longer than five minutes, and they always get this message. Anything under five minutes is fine and fast.
maybe the file is too large...
I can find the file, but it is 131 MBs...
@left citrus there is probably some limit on the file size?
@dapper pelican I will pass this on to folks on the team
sry
I tried the same track where I “got whole song” at 6:00, but that clip doesn’t DL either. And FYI, it’s not just this one track, I’ve had same problem with other stuff that’s shorter, but over five minutes.
Thanks.
The limit is WAY above that. I have downloaded a 27 minute one to test for example. Or the Infinity song. 300 or 400 MB
Wow! There seems to be a hard limit for me on anything over five minutes… I’m a paid subscriber too.
This is all .wavs you are trying though?
Yes. MP3 is fine at any size, WAV is good up to five minutes, will not download longer than that.
I'm guessing it's some sort of simple thing relating to how Suno does the export
Since .wavs are a new thing
and isn't tied to like, file size or anything
It may failing on the browser side for some reason. Try in a incognito window maybe, no extensions
I can find the file on disk...
so it is just the browser somehow stopps large downloads?
even discord doesn't allow me to send it...
Don’t think so, because I download/upload larger files than this from Bandcamp and SoundCloud all the time.
I’ll try a different browser though just to check. Like I said, a smaller earlier version of this track at six minutes, which would be about 60 MB, that doesn’t download either as WAV.
Different browser is always good first thing to try
Are you on a VPN or unusual network connection or anything, like satellite?
No.
My guess is Suno side
I'm sure z will see it and go "dang" and fix it
It's just .wav is so new
Like I said, I uploaded a 300+ MB mix to SoundCloud, just the other day which took a while, but it worked 🙂
It's strange that is only effects this one person though. That makes me think browser problem.
Or maybe, like, the song itself somehow. But that can't be the case, since they report all songs over some size don't work.
Let me try downloading right now
Maybe it broke recently
It’s not the song, I have tried several other songs that were over five minutes, and they don’t DL either.
Not 3.5. All v3.
Yeah so that would be out then
I just downloaded 170 MB wav of 3.0 song
I can send you a six minute file of another track if you want to compare.
Over 5 minutes too?
Yeah, 15 minutes
welp
Damn, well, I guess it’s working, just not for me…. 😦
User specific then
After clicking download, nothing happened for maybe 60 or 80 seconds.
Then it finally downloaded
I mean, it could be like one specific server/CDN or something, but it seems more likely to be something on the user side if I had to bet
What browsers did you try?
Long clip, I usually get the error message after about three minutes.
5 minute song appears faster than 1 minute for me. I can't explain that. I guess maybe some servers are faster than others, but kind of weird
Safari, DuckDuckGo…
DuckDuckGo may be an edge case, but I would be pretty sure Safari works. Still, you could try Chrome, just to match the browser I downloaded it in and rule it out for sure.
Safari is usually fine w/ big file sizes for me… do not use chrome, but will DL and try later. Will ask a friend to try too. I guess at least it’s good to know that some of you are getting larger files DLed. Thanks.
Update bugs if it works in Chrome. That would mean wavs are bugged in Safari, pretty big bug actually.
Will do.
Not sure if it's mentioned but downloading a wav seems to download about 10 copies. Very annoying.
Several people have reported that, but strangely it doesn’t happen to me in Safari….
interesting. i did see a chrome multiple download allow and thought i had other things. thanks
So the bug is whenever I download a .wav of the audio it keeps automatically downloading new files immediately after the preceding file has completed downloading.
yes same here. wish i didn't accept multiple downloads. sure there's a fix in chrome but shouldn't have to. i'm sure they will fix that.
Those bugs seem like they could be related...
A browser may sensibly abort a download if tons of repeated downloads are being triggered
the main thing they need to fix for me doing Deep House trax is the reverb on the strings and vocals get more and more with each extension.
that was the same in V3 as well
in time i think they will surpass Udio
Yeah, everybody’s saying that about the vocals, but I think that’s gonna take v4…
Although DLs under 5 min. in Safari show up fine and only one file.
@dapper pelican I can fix it mostly in my DAW with denoise and dereverb but would be nice not to have to.
DL failed in Edge & Ecosia as well.
I hear ya!
Friend was able to DL that big file in Edge quite quickly. I still can’t, which is weird.
You could try a VPN, maybe it is somehow the ISP, or the CDN you happen to be using (which would change with a CDN). Weird though.
MS OS for her, Mac OS for me…
At least I got the file now, lol. I’ll try a few things and see if I can ID the problem, but I would say keep an eye out for other people reporting the same. Oh, she did get the multiple download problem on Edge…
I'm using Opera, it was blocking the .WAV's, but after allowing them it all came good. Maybe the same issue with the other browsers?
But I can DL wav, just not Suno over ~ 50mb.
On v3.5, I’m getting a reverb / feedback / drone sound in the background for large portions of songs. It tends to have a synth-type sound, and can last through an entire verse and chorus, or even longer. It’s like a synth player hit a chord and is supposed to have a melody, but just holds the one chord forever. It’s happened on every song I’ve generated tonight, like probably 50+ generations.
That's is a problem in 3.5, but, it doesn't generally happen in every single generation. You could be prompting in a specific genre where it is extra bad, there are some. But even in that case there should be prompts that don't do it every time. Maybe try using Suno Search to find other songs in the same genre and alternative prompts of the same style.
The other way to deal with it is by listening closely to the songs and using Extend From to choose times before the effect kicks in. This is annoying but can usually avoid it.
Thanks, good to know it's a known problem. I'm just generating the same song over and over, adjusting the prompts each time. It's probably something in the the pop-punk, emo, metal realm that's doing it, because those have remained pretty consistent.
On 3.5, after 3 minutes of singing, the voice becomes lower and lower and then disappears among background white noise... quite annoying
While I greatly appreciate the possibility of having 4 minutes generations, I already had a couple that lasted 30 seconds, and I still got billed for a whole generation. v3 would generate 2 minutes clips quite consistently. I would like to be refunded for those bugged 30 seconds generations
also, putting a "max duration" as an option (once the above gets fixed) would be great, maybe even something simple like (long - up to 4 minutes, short - up to 2 minutes)
This might not be a bug, but if 3.5 can make longer songs, wouldn’t it be worth accepting more text as an input?
Reuse Prompt on a continuation should also reuse the Extend from clip.
i have been using 3.5 its crazy that i miss v2 even with all its faults
@solar grove
v3.5 does not listen to [intro Instrumental] or [intro bass guitar slap solo]. It just starts straight at the lyrics.
created 15 songs with;
[Start]
[Intro Instrumental]
[bamboo flute-solo]
None of them started with instrumental, they all started with someone breathing and the lyrics started right after.
https://suno.com/song/6fe58054-4f6a-4282-b3f2-43b2b6931a82
I found that the music structural always have problems after 2 mins
heartfelt emo, classical pop, piano, 125 bpm, female vocals song. Listen and make your own with Suno.
I've gone through at least 18 creations today for v3.5 and each song has the music style description being spaztically inserted randomly throughout the song. And the worst part is that they are all great songs otherwise
I'm sticking to v3 for the time being, lost too many credits on <1min generations or otherwise "dirty" tracks
If you haven't already, be sure to downvote v3.5 generations that have bad audio quality. (This should help them with improving it)
If I use Instrumental (V3.5), one of song always only 10+ secs
look like this one
https://suno.com/song/c3b92bb9-ad36-4062-93c8-589cd119a11d
instrumental cinematic classical darkness soundtrack cold mysterious gothic rock song. Listen and make your own with Suno.
I'm doing that but I have to say that reading "4 minutes generations for Pro users" and then seeing so many <1min generations is bordering the insult
since we're all alpha-testing 3.5, price structure for 3.5 should be different
or even zero
what I saw is that the short generations problem is more prominent for Instrumental tracks, still I think I should receive a refund on credits wasted on generations even shorter than v3 did..
medieval vocal monophonic song. Listen and make your own with Suno.
I'm not sure if anyone else has experienced this? I've noticed recently that the longer a song goes the audio quality drops. This song was an instrumental I made off the intro to another song that I like. By the end of the song the guitar that started out beautiful ends up very tinny
https://suno.com/song/8afffe89-67e2-48cc-ac81-9ca22f46d4d6
Mexican Ska, Upbeat Mariachi, Spanish Dance, walking bass song. Listen and make your own with Suno.
It happens quite a lot. The only workaround for it is to extend from before the quality drops.
Glad to know I'm not crazy 🙂 Hopefully they'll get it figured out sooner rather than later. Thanks for the reply.
Ye, suno should increase the quality of song, to much credits waste because this noise song in background destroy everything and to much time
And that, again, will cost credits, for something that's not my fault
Are we customers or are we (paying) testers?
Thx for the hint, for me it was mainly between 5k and 8k.
I'm just trying to help by suggesting a workaround. 🤷♂️
getting constant crashes
ie. credits refunded and no generation
i'm trying to extend a song
Suno don't follow my clearest instructions, i want male voice on verse 1 and 3 and female voice on 2 and 4 : (pegi 13)
[instrumental intro]
[verse 1 : male voice]
Ma douceur, je n'pense qu'à conquérir ton Coeur,
Ma douceur. [pause]
Ma douceur, je vais te montrer ma valeur.
Ma douceur.
[instrumental]
[verse 2 : female voice]
Mon bébé, lâche ce porte monnaie.
Mon bébé. [pause]
Mon bébé, je ne suis pas un projet.
Mon bébé.
[instrumental]
[Verse 3 : male voice]
Ma p'tite puce, je veux seulement qu'tu suces.
Ma p'tite puce. [pause]
Ma p'tite puce, verni moi le prépuce.
Ma p'tite puce.
[bridge]
[Verse 4 : female voice]
Mon chéri, tu penses qu'avec ta bite.
Mon chéri. [pause]
Mon chéri, et tu t'penses être une réussite.
Mon chéri.
[build down the instru]
[end]
Battle of rap, Duo singing, Couple, Female voice for verse 2 and 4, male voice for verse 1 an 3, R&B, rap, slam
i give up ...
I did too. I'll be back to Suno once it's proven it's reliable.
You're right and sorry for sounding so rude, I was just frustrated
And thanks
I am not sure if I reported this already fully.. or just made a meme post about it.
But I have extended this song (V3):
https://suno.com/song/35a6ab8d-0ef4-4b81-a212-a48e57e8e655
With V3.5: Wich game me this extension:
https://suno.com/song/6714db02-fec1-4c06-9bd4-f26402a26b16
Which gave me a full clip which was silent after ~00:43
Curious how I am... I extended that clip again and got a full 2:00 extension of silence:
https://suno.com/song/01f75782-17d5-4f08-9383-8c49c4baebd6
Here is it all merged together if that helps (history wise) silence after 7:47
https://suno.com/song/f6a9ae1f-5027-440e-a6b0-50d4fdf0345d
why can it be like this? Is this a bug or is it just my internet connection?
You are continuing from a part where the AI get's confused. So if you just change the time stamp 1 or 2 seconds, it will probably work better.
[Break]
[Instrumental Interlude]
Sad Shout:
Lyrics, Lyrics, Lyrics,
my format lyrics, its extend from songs
is this format correct?
Please read my answer above. I's more the time stamp, the spot where you are continuing from.
i want extend from 00:51, Because the lyrics are too tight, I want extand to add some instruments
Sometimes you have to go with the flow though. Even if 0:50 still has some lyrics, it usually finish the lyrics.
are you sure? this a POV
Yes.. many of us experience this. It's a common thing. And if 1 second back doesn't work, you have to go 2 seconds back. If it's eating part of your lyrics (not adding them), you will have to add the last words it's eating.
Sometimes you drive the AI, and others the AI drives you.
Been using the v3.5 beta and this is what I've noticed:
It struggles to adhere to genre prompts much more so when compared to V3. I can say I want a slow jazz song and it'll give me a modern country song. Frustratingly, the genre will often switch by the 2nd verse - so what might start out as a traditional folk song or 1950s doo-wop will become an EDM hybrid or 2000s style ballad.
Also in 3.5, the voice quality gets worse as the song progresses. Part of this is a solo voice will usually get multiplied for a chorus (even when it's not appropriate for the genre or the prompt says solo) and then stay that way for the rest of the song.
I think this applies to all versions, but when choruses are repeated they often don't have the same cadence/flow/enunciation/etc as the previously established chorus even when the words are the same. When the chorus is part of an extended track it might not even have the same BPM or sound anything like the rest of the song.
For a couple months now I've been seeing a UI state issue with the Instrumental toggle. It might show off but when I click create it returns a couple instrumental tracks instead. I have to retoggle it on and off sometimes.
have you a prompt to make a parody of a french song " ne me quittes pas" ?
Why are you asking this in the channel which is meant to report bugs? I think you are posting in the wrong channel 😄
sorry
No problem. Questions like this fit more in #💬┃general-chat or #📚┃suno-school 🌼
Right so here is a consistent, repeatable thing with the v3.5a. I have a spoken word intro I like, here:
https://suno.com/song/b34fa33f-0be3-42d7-9108-1e867c828a8a
The lyric box is this:
`[Judge]
Answer the question, Mr. Dellavue.
[Mr. Dellavue]
I don't recall.
[Lawyer Woman]
Were you, or were you not, at Ms. Samantha's house on the night she was murdered?
[Mr. Dellavue]
We had a fight. `
I cut from 00:11 seconds as I want to change where the song goes musically, and that's a great spot. The next portion of the song with v3.5a has this contained in the start of its lyrics:
`[Break]
[Drop]
[Melodic Verse]
Answer the question
Answer the question
~~~~~~~!
Answer the question
~~~~~~~!
Mr. Dellavue.
[Melodic Chorus]
I don't recall
~~~~~~~!
I don't recall
~~~~~~~!
I don't recall
~~~~~~~!
[Melodic Verse]
Answer the question
~~~~~~~!
Answer the question
~~~~~~~!
Answer the question
~~~~~~~!
Mr. Dellavue.`
Without fail the continuation will smash phrases together in a way that, if I didn't understand the language, sounds good, but makes no sense. And it seems to do it in the middle of a phrase, with things like "I don't requestion" (mashing of 'I don't recall' and 'answer the question' even though one is in a verse and the other is in the chorus) or "answer the Dellavue". I can consistently get it to do this off of this starting clip continuing from 00:11:
https://suno.com/song/b34fa33f-0be3-42d7-9108-1e867c828a8a
And here are continuations from that doing the weird phrase mash-up thing:
https://suno.com/song/a86a779d-834c-4633-9750-763cf358a3d6
https://suno.com/song/5b3857fe-eb41-4fe8-ad37-7e0f68957bb5
https://suno.com/song/459c5d66-809b-4a9d-922d-cbe0fff8be9c
https://suno.com/song/635b09b3-b41d-463b-a554-96b57585e793
https://suno.com/song/128817c5-1427-4272-9913-6ce06b43a78e
It seems, specifically the
`~~~~~~~! `
lines are the reason it gets strange with mushing phrases together as if I remove all those lines. The reason I am using them is to space out the lyrics as they are sung as I can add space between the lines as they are sung with `~~~~~~~`! but it specifically seems to mess up the lyrics in this way. This something I do in v3, but the sensitivity of v3.5a seems to not allow this 'fill' without Suno reading it...somehow.
I can work around this and change how I prompt, but it *is* interesting
If i am not stupid i am missing ~1500 credits
Did i miss something?
I assume you are aware that credits refresh every month with paid plan? If not, that might be it.
part 6 ? you do 10min musics ???
i had nearly 2000 credits and unsubscribed. Now i logged in after a while and i have 550 left 😮
it can be a mistake by me
But they do not roll over from month to month.. They are only for 30 days
so they get deleted if not used?
yes.
The changes seem to affect all versions. Why are all attempts faded out shortly before the 2 minutes in Free Account? You can't create an extension, it would be pointless. If you enter German, for example, a German text is generated in the first part, but a manual text is expected in the extension, which the text generation makes an English text. If you switch to instrumental, nothing works either. All efforts do not help and no extension can be created. The first part is therefore unusable. So you have to start all over again and the circle goes the next round. The fade-outs also add strange noises that are not intended. Previously, everything worked with English songs. Should the free account be restricted? It's no fun anymore and there's no point in subscribing. Is Suno now alpha again?
You can't use 3.5 with free accounts. But I witnessed the same deterioration in the songs that were created after its introduction.
Ever since the songs lost their old pictures, it's been like this! Of course I can't use version 3.5. But since the new version was installed, some of the extensions are now in much poorer audio quality. I also had 3 different styles for one title. All in the last few days since Whitsun. Calculations are also very often canceled and usually credited, but not all miscalculations.
While I like 3.5, it seems like if the first generation goes beyond about 3 minutes, it accumulates a ton of high-frequency noise after the 3 minute mark, leading to a background "screech" or "nails on a chalkboard" sound. I've had this happen with continuations and v3 before, but not nearly as often as it's happening with the first generations of 3.5.
At this point, I'm just putting half the lyrics in and using "Extend From" to cut the generation before the noise starts, but ideally with a 4 minute max generation length, I'd like to be able to generate with all 4 minutes...
I get this a lot with v3.5 too. The post in #📢┃announcements says to downvote songs that have poor audio quality so please do this if you haven't already (and upvote any that have good audio quality - from the songs I've generated this sadly seems to be quite rare.) This is to help them train v3.5/v4. It's frustrating but extending from before the noise is the only workaround at the moment.
So... what do I do if it has good quality before the noise, but then accumulates noise? 😅 Seems like using it as a "bad" training example would also teach the models not to mimic the actually good parts of it..
I'm downvoting these (I never have bad quality from the start.) Basically if the noise becomes obvious for a big part of the song,
i cant download wav file? it was doing that yesterday too
Anyone else using v3.5 and notice the voices are all more robotic sounding and there is a constant static over everything?
This sucks because v3.5 is awesome but then ruins everything to the point I can't use any of it
Not sure what you mean by "static" but I find with both V3 and 3.5, it adds a lot of reverb or something that gets louder with each extension and when splitting the stems, the "Instruments" seems to show more of this than the vocals. Adding a DeNoise and DeReverb helps but takes away other dynamics.
i can not get drift phonk to generate at all it was working in v3 and doesnt seem to work at all in v3.5
same issue and its driving me crazy
v3 was working
3.5 just picks whatever genre it wants at this point regardless of prompt
"I have a problem, the song starts clearly and distinctly then it transitions to a dragging scraping sound with high tones, like cat claws in the ears."
seems that downloading as wav downloads the track multiple times (only happens on full songs not parts)
Using dashes instead of spaces for a compounded terms will help (e.g. Glam-Metal)
See Jonathan Fly's post about 3.5: #📚┃suno-school message
@fervent fossil
nope just more wasted points
had to use v3 to even get close
the strangest thing. instrumental only v3.5 gen... and it made... talking? like the sims... at least for the first few seconds before the music kicked in
the cover art never stays the same while extending the song.
i wouldnt mind it if it asked me if i wanted to change it
and what is mess?
150-175 BPM, medium paced beat, chill, pop, bingbong song. Listen and make your own with Suno.
not sure what went wrong?
I have a song which I've worked on for ages and spent thousands of credits on. I now finished it, and as I played the final merged version I noticed that the sound quality of the early parts of the song has degraded significantly through successive merges. Can something be done about this?
Compare for example chorus 1 (02:25) in these two versions.
Early version, good sound quality:
https://suno.com/song/4c21993e-a938-4b0d-a8ac-113b327ebea1
Late version, bad sound quality:
https://suno.com/song/e42e9e51-c6f9-4ba8-812b-5a59a8ea3a06
Ancient Rome, ancient Greece, lyre intro, primitive, bardic, high notes, Nordic viking folk rock, symphonic power metal song. Listen and make your own with Suno.
Ancient Rome, ancient Greece, lyre intro, primitive, bardic, high notes, Nordic viking folk rock, symphonic power metal song. Listen and make your own with Suno.
when we extend, it load the good lyrics and style, but it don't load the tilte, it stay the last tilte we used ...
I think some kind of bug appeared. smo vrti won't finish the song and he took the points
3.5 has no understanding of "Acoustic guitar", it's distorted https://suno.com/song/660c9970-fb36-4d17-9ffb-34ff80b567f3
Acoustic guitar, flute, ominious, mysterious, mystical, fate song. Listen and make your own with Suno.
Dear people, I'm sorry about the pin, but I really beg you, the quality at the end of the songs, please
If you are experiencing poor quality with v3.5 make sure to downvote such clips with the thumbs-down button (and upvote clips that are high quality.) This is what the developers will be using to improve the quality.
on mobile there's a bug that doesn't let me generate a song with lyrics longer than 1250 characters even though I have v3.5 selected. if I switch the app to desktop mode then it works fine, but in mobile mode it refuses to work
this is crazy , i have more 9-45 sec songs than actual full songs with 3.5 smh wasted credits
When everyone generates rock songs with version 3.5, does it end up with very noisy electric guitar overtones continuously drowning out the singer?
I've experienced this situation several times already.
the end hurts in the ears 2:10 - https://suno.com/song/e7382df6-6725-4232-9239-1e70e6bc9ba3
Ambient Pop song. Listen and make your own with Suno.
Don't know what you mean
wow!?!?! it sounds good to you from 2:10?
yea
ok, we probably have different pain tolerances. The instruments in the background are way off.
Probably yea
This one also about 2:10 good for you?
https://suno.com/song/d2c8777b-8f93-4287-8bfd-f03513106039
It doesn't hurt my ears but I can agree that the background music is abit too loud for my liking
Suno stopped playing audio for me. Is this happening to anyone else? Other audio a sources work (youtube,Twitter), but no audio from suno even though the file is playing
I'm on phone browser. Cant seem to find the option for that
works on my iphone (edge Browser)
I've closed the tab and opened a new one on on Chrome iPhone
Thanks. I went landscape mode and saw the volume control then. It was muted. Didn't see it in portrait tho
Why always this bad background noise, it's annoying, it sounds like emery cloth abrading metal....
it's a Beta. hope they fix it soon. just give the tracks a thums down
If the sound is bad, please give it a thumbs down 👎 . It will help the devs to solve that problem
The problem is not new, they never really fixed it, in version 3 it's practically the same thing
I am sorry. I was under the impression you were talking about V3.5
This sounds more like a suggestion. Double posting is against the #rules. So I will remove it here. But your message is loud and clear 😉 @sharp ermine
I was talking about version 3.5 actually
Why can't I download WAV format files for my version 3.5 songs today😂
Then my answer is valid. If you use v3.5, the devs really would appreciate your 👎 to locate the problem in that model and make it better.
For me it doesn't fail.
- Is it a long track in minutes?
- Is it on your phone or PC?
- What browser do you use?
It might help the devs
PC,Chrome,3min,The previous version 3.5 songs could be downloaded normally.
I've already done this, as well as pointed out the bugs that were bothering me, but it doesn't seem to be much use. I have been using Suno for about 3 months, and this problem is recurring and not fixed. It appears on certain styles or when the musical intensity increases, but you should already know all that..
So it's related to 1 song atm?
[Nu-jazz],Eerie-drum and bass ,Saxophone,[j-pop],[Experimental-Electronic],[syncopated anime],gritty-
Female-vocals song. Listen and make your own with Suno.
Just to be clear, I am a volunteer moderator. I am not a developer. I am just trying to explain you what developers tell us, and how to handle things like that. I personally agree that V3 was not mature, nor is V3.5. But this is the normal state of the product tbh. If you expect perfection, I have to disappoint you. Still much to be developed and discovered.
Thanks! that would be helpful for the devs 🙏
https://suno.com/song/c2d4d105-47aa-4896-a58e-8edb7932ac96 This one also cannot be downloaded in WAV format.😂
[J-pop],[synthesizer],cute song. Listen and make your own with Suno.
OK I understand. Without striving for perfection, we can always hope that things will improve, right? In what I am submitting to you, nothing has changed, except that my bank account has been reduced by $30...
I actually have 2 premium accounts, paid for a full year. So I am a paying user in that regard too. Devs often say it's not perfect. They've been honest about that. They would LOVE to offer us the very best. But that takes time. This is new tech, and imho nobody does it better than Suno. It's just a matter of time. And they are working hard. They hope to bring out V4 with inpainting in 4 to 5 weeks.
dunno if this is a bug or anything but if you try female singer in any form and anthemic then the anthemic will force a male voice. the moment i removed anthemic i got my female singer back
figured i let people be aware
Dear Suno Team,
We would like to extend our subscription; however, we wish to access the data through Google instead of the Microsoft account. We have already tried linking both accounts, but without success.
We are now concerned that once I log out, all data might be lost, which would be a disaster. Could you help us link the accounts meltinganthem448 and motivationalpyrotechnicians469 so that we can access them through the Google account associated with this email address: germandarkpop@gmail.com and extend the subscription?
Best regards,
You'll need to email support@suno.com to sort that out.
With v 3.0, I was able to get a final song output and continue extending from that later.
v3.5 breaks this process, as after the complete song output upcoming part gains a lot of volume, creating distortions and cracks in the sound.
Been using up my credits before my sub expires. Quick note to say as I have been, 3.5 instances were like a blind darts player trying to hit the board of the right genre, and never got it right. V3 nailed it after 6 tries. 3.5 sound degradation over extending was hard to listen to with noise drowning out vocals at times. V.3 was bad for this but not that bad.
I'm going back to 'free account' until this platform fixes its issues, but really looking forward to v.7 or 8. 😉
Not sure if this is a bug, but it's annoying af when the ai keeps making a quiet outro to end even tho my prompt is something like [loud energetic vocals] or [shout] and the ai generates a quote whisper. I've wasted so many credits and the ai just keeps doing this. I want the energy to carry through all the way to the end, not end in whisper that you can barely hear
Have you tried capslocking the words? Yeah the ai is dumb a lot of the time
Also on this note, I feel like the [Outro] and [end] tag have too much priority. Sometimes I'll try to repeat a chorus twice before going into a final outro chorus to end, but the ai ignores everything and just does a single outro chorus. Has happened consistently and have wasted prob over 1000 credits on regenerating songs the ai effs up this way
Oh. Lemme try this
Didn’t work sadly. Kinda getting annoyed. Getting really annoyed with this honestly. I’ve pissed away so many credits cause the just goes off the wall with the ending and does nowhere close to what I want for the end. It’s like everything is great up to the end, then bam 100 credits just to get it to end correctly
https://suno.com/song/fcb90c3f-7f1a-4d42-a0a6-27d0c84f0c48 like this for example. It's so quiet and muted
alt rock, indie, post punk, new wave, pop rock, synth pop, solo male vocal, energetic, upbeat, fast song. Listen and make your own with Suno.
And here it is with all caps https://suno.com/song/c8732b64-ad2c-4bf0-a1c5-f0fa4851df7f
alt rock, indie, post punk, new wave, pop rock, synth pop, solo male vocal, energetic, upbeat, fast song. Listen and make your own with Suno.
Maybe they can add a specific (ending) tag you can put in the music style boxes -- but alas, the two times I tried it my suggestion (probably a placebo) did work. The songs weren't good anyways though so I scrapped em haha. Best of luck man
And then it picks up again after fading out
.....
I think the issue is the part km extending from is a quiet part (which is what I want) but then I want it to pick up again for the ending
Honestly I am not happy with the way the vocals keep wanting to climb and distort will it sounds like a vocaloid on crack, no matter what I do especially with symphonic metal variations the female voices start clear and then just go off the rails, even trying to prompt for lower vocals like a bass is useless, I hope they get the vocals fixed cause I can not waste credits on generations I can not use
Trying style like "Greek influence" or "Arabian influence" which always gave Greek/Arabian unique sound in V3 doesn't work at all in V3.5, I guess for other styles area related. also V3.5 having hard time reading prompts. "guitar solo" didn't give me one single solo.
Finally got it. I added a [Chorus 3] tag and finally did what I wanted. Had to spend another 30 creds to get what I liked tho
Yea. The vocals get really distorted in v3.5 for me. It status to sound very robotic and artifical and quality degrades exponentially
Cancelled my subscription a little while ago and now a month after my last payment, I've lost the remaining credits I had and am now back on "trial" credits.
Can you explain what the problem is then?
That I paid for those credits, and the subscription, I'd assume that those remaining credits (at least 1.5k) should have remained in my account
Did i miss something about credits being deleted after a subscription runs out?
I guess you didn't fully understand then. Credits do not roll over from month to month. Just like a most phone plans.
where is this exactly? I've been searching through everything but can't find it anywhere
Subscription page.. where you also choose the plan
Scroll all to the bottom
Ah, completely read past that and went straight to the TOS page
Here's another example of what Kate Suno mentioned about the quality decaying towards the end. v3.0 had this effect too. https://suno.com/song/eb741aa0-1897-47b6-909f-de14ace81ba2
Bossa Nova, Jazz, Samba, female solo singer song. Listen and make your own with Suno.
And yes, i'm downvoting the bad generations, but I thought a more specific example + description of the problem was probably more informative than just some unspecified downvotes.
i cant gernerate songs why???
Check if you still have credits.
no how can i get them only for paying?
You get 50 credits every 24 hours. Refreshes of credits will be at 0300 UTC+0 (<t:1713236400:t> in your local time)
Or you can buy credits, yes.
ok ty
oh btw, i'm getting slightly better voices with v3.5 than v3.0, so that's nice. it's still very far from the quality i got from v3alpha though.
The sound starts to degrade around 0:58, very similar to the results I have been getting since v3Alpha.
yeah, and i didn't cherry pick this specific song to showcase this problem, it happens to the vast majority of what i generate. in fact, this was the good take that i kept.
i downvoted 15 bad ones.
I agree. I have to extend music with model 3.5 before 00:45. Distorsion and « hallucinations » becomes too audibles after this timestamp.
For song I use only v3
Let's have a few random v3alpha generations for comparison, in case someone here joined after the switch:
https://suno.com/song/e705f5f5-fd29-465b-b2b5-9ffa29f13edd
https://suno.com/song/681c0177-2504-4811-9660-a87e14b32e0b
https://suno.com/song/636de206-e8d3-4f23-bef5-4b26e9892b82
https://suno.com/song/0056eaef-b6d6-4e21-bb53-9f0b8744a3e7
i just dug these up just now, in case someone asks "was it really that much better?"
It will be good that Suno offers extra credits for Pro account who use 3.5 model in this « early beta access »
why is it a weird reverb middle of all 3.5 songs ? and vocals get real crunchy halfway through
yeah, and the reverb never goes away once it's there
so the first half minute is usually nice, but then it just dissolves...
like putting a nice song into a jar of acid 😄
yea bro this is crzy. so much wasted credits
now i gotta remember to keep my lil 30$ next month
that fact that it did work earlier is probably what makes it a little bit frustrating
having something taken away from you is always worse than never having had it in the first place
i should go write a blues song now...
I want to wait until we have inpainting, I want to have the vocal part of this, but I want to put it between the other extensions from this track as a whole song: https://suno.com/song/bc9406b1-f51e-459c-bb04-3501e18f8345
Alternative Rock song. Listen and make your own with Suno.
I do wonder what changed between v3 alpha and v3 which has been difficult to undo. It can't be that simple though or they would've dealt with it by now.
that is indeed the great mystery!
admitted, v3alpha was not as good at following [cues] and getting the vocal structure right as v3.0 and v3.5, but that's a limitation i could easily live with.
i also think it was less prone to favoring pop music
I like what I'm hearing in v3.5 early access, but still has lots of 'noise' in the background... and sometimes near the end of the songs will just be more noise than music.
This is a known issue and developers have asked for users to "downvote" or "dislike" using the thumbs-down button to highlight generated songs that do this.
Aanon: it has almost as much noise as 3.0 ... it's quite similar
I was hearing it more than v3. Maybe the music I was picking. Happened more with "instrumentals".
oh okay
seems like Dutar has slipped out of the training at some point, i noticed it on 3.5, but I'm not sure its still in the 3.0 model either as the results I got seemed more like harp/mandolin (actually, seems to be the case for all uzbek instruments... cant remember which ones used to be trained, but I had used several in the past)
I downloaded a wav version of a song and clicked it once, a minute later I got 7 downloads of the same file.
I tried it again on the same file and only got 1 download this time, then tried a different song and got 4 copies.
It has probably been reported already, but thought I'd toss it in here just in case
I'm really enjoying the updates! Thanks for the excellent product! One thing I'm noticing is that when extending using v3.5, about 75%-90% of the time it's filling the entire two minutes with lyrics that weren't in the extended prompt (mostly repeating chorus from previous prompt several times). And the majority of these extended songs have are having a hard time following the [end] command after my [outro] section and are continuing past the 2m mark, just seemingly right before the perfect ending lol. Just wanted to share that feedback. If you'd like anymore details, please let me know. Over than that I'm loving the new limits and outputs from the new version. Thanks again!
V3.5 + Acoustic Desert Nubidian Acoustic nu-metal, Dark Electro-steamfunk Dystopian - Is giving me lots of seeds that have singing 😄 (as instrumentals)
is there a max limit to your trash bin? i'm at 14 pages, and i deleted some songs, but i wanted to recover one cause i deleted it by accident, but i'm not seeing it or any of the new ones i deleted
I'm having it happen on all generations using 3.0 and 3.5 right now.
Running into a few bugs: 1. Suno repeating a verse instead of moving forward with the lyrics. 2. When I then go to extend these songs I choose a timeframe where I want it to start from and sometimes it brings in the repeat from previously.
happens when the syncopation is weierd and its trying to move forward - unfortunatly you may have to cut and extend mid chorus if its super bad
Yeah I have done that unfortunately a few times with cutting it and extending.
you can always try to prompt multiple times [outro] [outro] [end] helps in general I think
sry for this -- typically it is really lyrics dependent....
like Furrybeans said, downvote is the best option now
a trick that i've found to end a song when it doesn't want to is to extend and put nothing but [end] in the prompt. it usually creates 3-5 seconds of silences, and u usually get refunded for it when it's that short, u can then use that to complete ur song
v3.5 is terrible for EDM/house music.
I've used up 700+ of credits and almost all of the EDM tracks it generated didn't have a clear beginning and became noisy or "scratchy" as the song continues.
v3 is still best for EDM songs.
It seems that the fuzzines increasing during the song is a general issue with v3.5 ... it's verious obvious on human voices
Is there any general rule here or specific to outro. Like if a metatag does not run as expected, put it multiple times.
I just use [Fade to End] or [End]
I discovered an interesting bug. It has to do with the site. When I was listening to a song I refreshed the page and when it popped back up my songs were all gone except 1. After a few minutes I came back and the song had changed to a completely different song of mine that vanished. So maybe the songs are just hidden and it will fix itself. Either way it is a little frustrating.
I use [Fade Out][Song End] and it works most of the time. Sometimes it starts singing again but then I continue from a bit earlier and try again.
I haven't gotten an extension under 38s using v3.5 even with only [end] in the prompt. Most go on a bit longer. But still very happy with the updates overall 🙂
fuzzines increasing during the song espesially towards the end of song. Guitar becomes just a scratchy sea of noise no distinctive cords is a general issue with v3.5. I also notice but i may be wrong the songs souds the same all mellow tone
I'm very happy to have recieved an email congratulating on 100,000 plays on a song, though the thing is my current highest song is 28,000 plays 😆
So I believe this is a bug of some sort 🤔
Did anyone else get this also??
yes, I heard more people with a simular story.
v3.5 is usually the voice comes from very very deep, like another dimension kind, very muted, and fluctiating most of the time. And since the afrival of v3.5, now v3 started to ignore any prompt at the beginning of the song, and it will almost all the time immediately start singing the fiorst line from 00:00 being the start of the song, that immediately. noi matter what promts i gave, it would either keep it extremely short or it just totaly ignores and get on with singing just like that.
anyone havonbg similar?
I've discovered that you can get longer instrumental tracks out of v3.5 if you don't use custom mode, but while doing this I generated a song that has a style is longer than what you can enter into the style field in custom mode, so I can't extend it without changing the style.
https://suno.com/song/f0a83110-af9b-4250-8d62-4aa1ea7d216e
total modulation futuristic post-rock energetic electronic industrial trip-hop panning dark mysterious gothic space opera 16-bit yamaha ambient downtempo g major dueling leads song. Listen and make your own with Suno.
chirp-beta
Chirp is gone, you cannot make music in discord. You can make music on https://suno.com
[Instrumental Intro]
(heavy guitar riff and pounding drums) this is leading that Suno is generating vocals singing "heavy guitar riff and pounding drums" in version 3.5
Things in (parentheses) are intended to be sung.
Things in [square brackets] are not.
It's for when you want something to be repeated like by a second/backing vocalist, e.g. "take me higher (higher)"
I know, but it was generated using simple mode...
I did not put it there myself
Another one: with simple mode there are only 6 'structural' parts being generated. i.e. verse 1, chorus, verse 2, chorus, bridge, solo and then it stops...
I'd like to have more without extending
extending=sound loss
There appears to be a bug of some sort with the cover images.
I'm on the free plan and, if I use Edit Details on one of my full songs it brings up this.
It allows me to select one of the other images (from the clips which make up the full song) and click Submit. A message is displayed saying "Song details updated successfully." and the song cover image changes, but this is lost when refreshing the page.
Either users on the free plan shouldn't be able to do this, and the option to select a different cover image is being displayed when it shouldn't, or the changing of the cover image is not sticking.
thanks, working on fixing this
I have reported 5 different instances where the output was a completely different result than what was requested. Will this be reflected somehow? The style, the genre & song name will be something completely different. Im sure I'm not the only 1.
But will I get those attempts back or I just got dealt bad cards?
I don't think you get anything. Not even the subscribers. Although I'm not sure there cause I'm on the free plan.
i'm finding a lot of issues with v3.5:
-
if you extend from a region/point where a word was just ending some lyrics are skipped, singing style is often very weird/unnatural, or the resulting clip ends up with errors.
-
when extending at certain instrumental parts, the following lyrics are often sung in a very unnatural style.
-
extending with no lyrics, puts gibberish in there or repeated vocals.. here the expectation is that it should continue as an instrumental.. this makes it hard to extend parts where extended instrumental solos might be needed, for example, at the end
it would be nice if they 1) add more meta tags; and 2) consider everything between squared brackets as instructions .. this might allow people more control over the style, etc. .. i already see this being better in v3.5
So I've wasted over 200 credits trying to fix this, but v3.5 gets stuck repeating the wrong chorus and no matter what I do it won't correct.
I wrote two chorus, and I wanted them to alternate between the two different choruses. But it keeps using only the 2nd chorus, and despite how many times I've extended it using the lyrics from chorus 1, it only songs chorus 2 instead
And it's frustrating me to no end that im pissing away credit alike this. I've extended from diff parts of the song, replaced entire verses and bridge, but it still refuses to sing the correct chorus that's actually in the lyrics
i spent more than 2000 credits doing the same 😆 in the end, i just made use of the clips i had and edited them myself
Yea. At this point prob export it to some outside program to fix it
Hopefully they’ll let us do this on the site in some future version
Yeah the Heavy Metal Section is pretty much doomed hard to get good songs#
What's even more frustrating is it did it correctly for my original generation. But then I tried to Change the genre and female vocals and it's just effing up majorly
Also encountering a generation error when trying to swap to v3 to extend from a v3.5 song. Was trying to see if this would fix the ai but guess not
Your problem is most likely the length of your lyrics. V3.5 can handle double the lyrics. So just remove some lyrics.
Hmmm. Ok. I'll try with a single chorus and see what happens. It's just that it worked for my very first generation. But i guess maybe cause it's beta
But then again. I think the text limit was in place when u first tried. Will give it a shot and see what happens
i also encountered another issue, that if you add lengthy lyrics, it speeds up the vocals almost making them sound cartoon-like in most cases and leaving less gaps for instrumentals/solos
EXTEND works fine within 3.5. No need to switch down to 3.
Nope. Another 100 credits and didn't work. Tried extending from diff parts. And then tried with just the four lines I wanted. It kept repeating the old chorus instead of the four new lines I wanted
Not really seeing how it's not wanting to read my lyrics at all
@wintry tundra If you want I can try to look at your song. Not sure if you want to share your link publicly, but feel free to send it to my DM with the lyrics and I can try to see if I can find a workaround
That rarely happens to me now in 3.5. But it happened to me yesterday. I fixed it by starting the EXTEND a few seconds before the end of the previous. (But thats so unintuitive it may not always fix this issue)
I'll just try again in a Few days after my credits reset. Might be stuck in some loop cause the first half of the chorus starts off the same in chorus 1 and 2, but then changes in the 2nd half (and even in the 2nd half, they both start with the same word but have diff words after that)
Maybe being in a mix of Korean and English is messing with it too. Thanks tho. If it still doesn't work in a few days I'll reach out to u.
Alright.. 🌼
I completely replaced the second half with completely diff lyrics and it sang those. But when I tried again with the original lyrics I wanted, it went back to the wrong chorus
extend doesnt work for me properly either .. tried with many variations with over 2000 credits .. it's quite messed up in 3.5
perhaps it's only glitchy in certain genres?
Last night I was working on a song in the create area on the web site. I had about 10 songs and then this song in that tab. I ended up refreshing the page as I thought that may help clear up the distortion I was constantly getting. When I did, the 10 songs vanished and only the one song I was working on remained. I noticed that at least 2 of the songs that were in my create tab were now just under my Liked area on the home page. They aren't in my library. One of the times I came back to the create page the song had switched to a different song that I thought was gone. So I think the songs are invisible on the create page. Is this the correct area to report this or should I be sending an email? If so, what is the email? I could only find the billing email.
Yea. The ai is stuck in some weird loop for my song. I had completely changed the chorus part and it sang thst, but when it hit the next part of the chorus with the lyrics I wanted (chorus 1), it defaulted back to singing chorus 2's lyrics instead.
In other words, it sang the new chorus 3 I made up as a test, but then when it hit chorus 1, it sang chorus 2 again instead of what's actually written
yes this happened to many several times as well.. from what i could gather, it has something to do with the point at which you try to extend
Ive tried many diff points aside from just starting firm the very beginning
Well. I haven't tried t starting from before chorus 2 begins yet
try different second mark around the same position, you might get a different result
I really like how the ai sang verses 3-4 and the bridge after chorus 2, but in afraid it'll change that melody if I extended from before chorus 2
Yea. I've tried 1 sec before and 1sec after
And I've also extended before the bridge (and it completely changed that melody), but still repeated chorus 2 after
And I've also tried frim like a min after my original extension point. Was trying to find a compromise point for how many times I would let chorus 2 repeat. But it just won't sing chorus 1 no matter where I extended from. Ie 2:16 > 2:26 > 3:26 extension times
it does that to try and find the right style of singing .. that's what i mentioned in my post a little bit before yours .. that it either repeats lyrics or spews out unnatural sounding lyrics
Might need to just start from the original chorus 2 point and hope the new melodies for the verses are still to my liking
Chorus 1 and 2 both share the same first 4 lines in Korean.And then a space after those lines, then chorus 1 repeats "I'm hot you're not" x3 in a mix of English and Korean. and chorus 2 repeats "I'm the light (I shine), you got no light" x3 in a mix of eng and Korean. But it keeps repeating the "I'm the light chorus" only and won't sing the "I'm hot you're not"
But yesterday when I was doing this, it worked just fine. But today when I tried it with diff genre styles, it's giving me this problem
i used english for the entire song so i doubt language might be an issue .. but either way, i hope they fix it in v4
for me v3.5 has so far not been a good update 😕
I've encountered an issue today. (this evening)
Previously I've been able to download the .WAV file but when I've been trying over the last hour, I just get a Download failure error please try again.
But I can download the .mp3 of the same files.
Don't know if it's me / just something temporary? or an issue?
--
Update (90 minutes later) - WAV downloads are now working again for me.
(1)Last night I was working on a song, I decided to refresh the page. When I did 10 other projects I had sitting on my create page vanished. Only the song I was working on remained. After several refreshes and looking around the site, when I came back to the create page the song had changed. (2)I noticed that in my likes, 2 songs I was currently working on are sitting on that page. Upon going to my library, songs I had reported as bugs were there and only one song I had been working on was in there. (3) I clicked on the liked tab to see if any were hiding in there. (4)When I did I guess I confused the bug that must be on going as one of my songs reappeared in my create page. At the time of refresh I had my current song that I was working on playing. @left citrus
The quality of the genre phonk is extremly bad
I think I mightve found what went wrong. So what happened is the second chorus got repeated twice and I didn't realize, and I think that caused the ai to get stuck in a loop. Song was supposed to go verse 1, Verse 2, chorus 1, Verse 3, chorus 2, bridge, chorus 1, Verse 4, chorus 2. But what happened was iaccidentally put in the prompt verse 3, chorus 3, chorus 2, bridge, chorus 2, verse 4, chorus 2.
So I think have chorus 2 repeat 3x messed up the ai's algorithm. So it just kept singing chorus 2 no matter what I put. So I went back to the bridge and replaced chorus 2 with chorus 1 for the proper order it it seems to have fixed it for now.
Now to spend a ton of credits generating a melody flow that I like
Previously I was trying to extend from aftet Verse 4 in order to preserve the melody and flow, but it wouldn't sing the correct chorus from anywhere after that. But I guess that's a sacrifice I have to make to fix the chorus.
Yea. I've been enjoying 3.5 so far, aside from the loop (ba dum tsss) it threw me for this morning lol
Yeah getting similar results on my generations, did two generations(so four total) Same lyrics for both, all 4 generations looped the first Verse, considering i can Snip them in another program the first verse in commonly the better one as it sounds like the start of a song whereas the repeated verse comes in like it's the second verses turn 😛
V3.5 does not know what reggae means
@wet lotus Blend it with other terms that also relate to reggae music 🙂
It's like a way of giving the AI Adderall focuses it in 😛
Does anyone really know?
The first thing I try in cases like that is use reggae, reggae, reggae, reggae instead.
Currently testing 3.5s attempts at already completed songs i made on V3, just reposting the lyrics with the same styles.
Just reattempting a D&D Hip-hop Song I already made in V3.5 so far it has switched up styles 3 times in the first generation, Started off sounding like Macklemore then started sounding like Linkin Park then went country >.<
It works in this case
So from my tests since 3.5s release, repeating verses and it has trouble maintaining the music style.
@low cypress I still can't download the 3.5 WAV of the new song. https://suno.com/song/bc01e9eb-0b6b-444e-89ac-79742d56732f 😂
yikes.. that sucks. But I cannot not fix that. I am not a developer, just a volunteer moderator. But I know the devs look at this channel at least once per day, if not more. So i am sure they will pick it up.
@left citrus
weird, i was able to
for you it was timing out?
huh ok, i'll look into that
It's always been like this, only the first 3.5 song can be downloaded in WAV format normally, the later 3.5 songs created don't downloaded in wav😂
alright, sorry about that. i'll do some testing on this tonight
thx💜 Thank you and looking forward to your gifts🐰
v3.5 is great, now I can listen to up to 4 minutes of paperwork. Most songs are 1-chord per bar trailer park music. Please take us back to the beginning of April.
Same experience here
v3 is still available if you prefer that
Don't get me started with late v3... I don't doubt it is the same model, but they aren't using the same computing power, and that's undeniable if you know a little bit about music.
I understand your concerns, but it's important to base our discussions on accurate information. Not assumptions.
I really hope they can fix that site problem. Just my luck, 4 days into my 1 year sub. Sigh.