Be sure to check out the unofficial guide for some tips but there is still a lot of experimenting on what actually works. I'll be posting my findings as well soon! tinyurl.com/2edht883
Thanks for your useful tips, I started using udio 3 days ago and struggled a bit getting good results for some genres. But once you get used to udio, it's great, just published my first vid with udio audio today and have another one in the pipeline already. Never tried manual mode so far, will test it. Btw. your raven queen song sounds really cool! 💖🐦👍
You’re welcome! That’s just it, it’s not perfect but when you get a good generation it’s incredible! I’m quite enjoying myself! I’m actually going to finish that song today and post it on my playlist 😊enjoy and have fun!
The chorus trick doesn't always work. We don't all write songs with sections short enough to fit two of them in 33 seconds and tempo differences mean you really can't limit people to 33 seconds and think this is always going to work. It only seems to work with a verse / chorus combination in the same generation. Try adding a bridge and it has a seizure. It's not flexible enough for most people who write complete songs.
I hear ya but you can extend before and after the clip. So if you had verse-chorus in the first clip, you can extend before to create a verse. Often what I do is 1st clip verse-chorus, extend before intro-verse, then extend after and go from there. Bit of a hassle but that's what we have to work with and I've made quite a few complete songs. It's far from perfect and there is much improvements to be made but considering this is still new, it's pretty impressive. They've so made updates so that you can extend from a certain part however it doesn't work too well at the moment.
With lots of generations you can usually find something that works. Looking forward to trying the new changes. I think it has 2 minute review, clipping capability and up to fifteen minute songs now. Just need to see if all my credits have renewed. I'm definitely looking forward to future changes on Udio.
I've found this only consistently works when the chorus and verse are in the same 30-second section i.e. Section 1 (Verse+Chorus) Section 2 (Verse+Chorus) which actually wouldn't be the way most songs are structured, 30 seconds is too short for both. I believe it works because the transformer model looks at the last section to predict what it will generate next. If the chorus sections are separated by a verse sections i.e. 4 sections containing Verse, Chorus, Verse, Chorus then getting consistent results is much harder as the model doesn't always use the context of the first chorus to generate the second. Perhaps a larger context window like in text models will solve this.
@wizards-themagicalconcert5048 unfortunately, right now atleast, the only way to get a consistent chorus, is by cramping a verse and a chorus in each 33 second generation. Example: First generation (Verse+Chorus) Extension (Verse 2+Chorus) and so on
I love this channel, so much to learn and it's explained so well. I tried to create a song with the same melody and lyrics but in two languages. Unfortunately, I couldn't find a way to do it. Now, I have two songs with the same lyrics but different melodies. Maybe that will be possible in the future.
Interesting! can't say I've tried that but yeah I suspect it will be difficult to do at this time. It's still very early for this kind of AI but over time I can see more flexibility and consistency.
Very good tips! I've been having issues with the consistency when my songs extend beyond about 3 minutes, and I don't have the chorus appear within the last minute or so, I'm guessing it has to do with the memory it retains. Also ending the song in a similar fashion as it begins is hard.
Glad it was helpful. They did make some updates recently so that getting consistent choruses are better. I believe now it can look back 2min so it has improved memory. Will post an update soon
Thanks for the very helpful video. Do you know how I can add a custom thumbnail to my song? Currently it is just the generic red blotch but I see others have ai generated colorful images.
I'd really love to have the Suno option of extending a song from a specific point. It would make things so much easier! Currently, I sometimes get outros that I like, but that don't come to a proper conclusion, so that I have to add another outro (instrumental) to get the song to actually end - provided, of course, that I don't exceed the 4m22s limit.
when you extend you can activate "crop" and you select what part you want to extend. It's even better that with suno because you can set the beggining point too. With suno you can't extend before... I really hope it will change because that's really limited...
Sometimes I can't choose between the 2 choices (or more if you generate again) that Udio gives you, so I develop both and then remix in a DAW. Had some great results that way and it means you can extend beyond 4 minutes 22 seconds, if your song needs to breathe
Yeah same here but not everyone knows how to use a DAW. With that being said they made some updates including longer songs. Up to 15min if I remember correctly. Will be making an update video soon.
Who can tell? What to do if you have already typed your text. I like the melody. But you decided to change the text a little. How to edit it while maintaining the old melody?
omg thank you so much for this... I wasted half a day and 200 tries trying to get something consistent and decent. I also challenged myself to write something in Korean xD which I don't recommend (Their system is not suitable for non-english song yet)
It can be stubborn a sometimes and can take quite a few tried to get it right, the model is still very new. I was very curious how it was with other languages. Generally though, the automated lyrics are not the best. Suno AI does a better job, or even chatgpt like I showed in this video 👍
thx for this video. very helpful! i have seen some ppl use [verse 1] [verse 2] [verse 3] etcpp vs [verse] [verse] [verse]. does this have an impact of the melodys consistancy? Same question goes for a [chorus]. so, does udio add variety if i use differeent numbering? thx
You're welcome! At this point I can't say it works consistently, I've tried it with numbers and without. What I do know is that it seems the success rate of getting the same verse and chorus is more likely if you put it on the first clip, then extend before or after. I know it's not ideal but as of now that's what I've done to get some consistency. It's still very early so I expect improvements to be made to the model eventually.
Not possible at the moment however you could use the same tags and manual mode and get a similar vocal sound. Especially if you put in the style of a particular artist. The downside is it’s totally random and could end up using a lot of credits.
Actually to think of it you can use remix and increase the slider all the way to similar. Although I haven’t tried it myself it could work, curious to test it myself now 👍🏼
You mean in places where it's not meant to? It does happen every now and then, you can increase the "lyric strength" in advance settings but not too much otherwise it will sound artificial. It's just the nature of the model at the current state. This will improve over time.
@@MonzonMediaLike I write my own lyrics so I may change something and it will keep saying the same verse that I changed before I changed it. Is 60 good for the lyric strength?
Привет! Спасибо за видео. Udio очень классный продукт, я повеселился, когда делал свои песни. Но, как только мне потребовалось сделать что-то сложнее, я потратил порядка 200 генераций, чтобы закончить песню. Приходилось прибегать к различным ухищрениям и редактировать текст таким образом, чтобы AI могла его уместить в размер и правильно расставлять ударения. Чтобы куплеты и припевы звучали единообразно я использовал такие пометки как [verse same as previous verse]. Udio выглядит как перспективная AI и, надеюсь, создатели усовершенствуют её. Сейчас с ней сложновато работать, т.к. это огромнейший рандом, а хотелось бы больше возможностей управления.
Doesn’t hurt to try to put into the tags, some claim it works but from my testing it’s just random. You have better luck using upbeat, slow tempo, aggressive things like that.
I'm trying to follow your instructions for replicating the chorus, but whenever I turn on Manual Mode, I get an error saying my generation attempt failed moderation checks, and no matter what I change, I still get the error.
@@MonzonMedia Still an issue. It's odd because the lyrics were auto-generated in the first place, and if I turn off manual mode it works with the same lyrics. I guess that's why they call it a beta? 😆Thanks, though!
Man, that is a very powerful tool. Have any idea of how to make money with it? I think making customs songs for birthday parties can be one option. What you guys think?
@MonzonMedia A copyright can only be put on a part that a human created/edited. No one can own the part that the AI created. -Top Music Attorney. Check her channel.
Thank you author for your work! But avoid lengthy, unclear expressions, and especially ambiguous ones. You have to re-listen several times to exactly what you meant! Express your thoughts in military terms, this is a program after all! Not a fantasy poem! The FAQ of the PROGRAM is also written in a florid, lengthy and not specific way! We want specifics from you! Good luck!
if you can keep the 2nd verse and the chorus within the 33 seconds on the second clip, then yes u will get the same chorus. but if your 2nd verse fills up the 33 seconds, and u wanna get to the same chorus after that, say goodbye to that chorus. it wont repeat. ive wasted 100s of credits. if anybody know how to get the same chorus again, pls tell me. p/s: i dunno if it it's better or not, but i generate the chorus first, then build up from there. i always start with chorus, then add before, add intro, then add post-chorus, yada2. then outro. i find chorus first, then work my way with the verses, etc. unless u add [intro] [verse] on ur first gen, then maybe it'll generate it properly.
Yes as mentioned it depends on the tempo+lyrics. It’s currently a limitation of the model and I suspect it’s how they preserve the quality without it going off the rails. Sometimes it will take several tries that’s the nature of generative Ai.
Not as impressed with this as I first was, lyrics are clear, but the music all sounds like punk with monotone vocals, clear, crisp, monotone vocals, that being said putting [spoken word] is interesting in an all instrumental
Remember this is still a beta model and even with Suno AI, it can be very random. Might take a few tries sometimes. It's also free beta at the moment. There is still a lot of experimentation going on, even the developers can't say for sure what works.
I use it to make the lyrics or get an artist...i make beats so i manipulate the hell out of it and im shocked at the results...I just signed my first AI artist to my new imaginary label
i got a longer result by adding intro last. it messed up the filename and wouldent make a video but i could dl audio and cobble this together w yt movie maker. th-cam.com/video/ci27ChXhq78/w-d-xo.html
Hey everyone, there have been some updates as of late, one big one is an increase of 15min max length. Update video to come!
Be sure to check out the unofficial guide for some tips but there is still a lot of experimenting on what actually works. I'll be posting my findings as well soon! tinyurl.com/2edht883
that link s very helpful! thx alot
Thanks for the tips! This AI is truly amazing!
Thanks for your useful tips, I started using udio 3 days ago and struggled a bit getting good results for some genres. But once you get used to udio, it's great, just published my first vid with udio audio today and have another one in the pipeline already. Never tried manual mode so far, will test it. Btw. your raven queen song sounds really cool! 💖🐦👍
You’re welcome! That’s just it, it’s not perfect but when you get a good generation it’s incredible! I’m quite enjoying myself! I’m actually going to finish that song today and post it on my playlist 😊enjoy and have fun!
The chorus trick doesn't always work. We don't all write songs with sections short enough to fit two of them in 33 seconds and tempo differences mean you really can't limit people to 33 seconds and think this is always going to work. It only seems to work with a verse / chorus combination in the same generation. Try adding a bridge and it has a seizure. It's not flexible enough for most people who write complete songs.
I hear ya but you can extend before and after the clip. So if you had verse-chorus in the first clip, you can extend before to create a verse. Often what I do is 1st clip verse-chorus, extend before intro-verse, then extend after and go from there. Bit of a hassle but that's what we have to work with and I've made quite a few complete songs. It's far from perfect and there is much improvements to be made but considering this is still new, it's pretty impressive. They've so made updates so that you can extend from a certain part however it doesn't work too well at the moment.
With lots of generations you can usually find something that works. Looking forward to trying the new changes. I think it has 2 minute review, clipping capability and up to fifteen minute songs now. Just need to see if all my credits have renewed. I'm definitely looking forward to future changes on Udio.
Super video. Thanks for posting!
You're welcome!
I've found this only consistently works when the chorus and verse are in the same 30-second section i.e. Section 1 (Verse+Chorus) Section 2 (Verse+Chorus) which actually wouldn't be the way most songs are structured, 30 seconds is too short for both. I believe it works because the transformer model looks at the last section to predict what it will generate next. If the chorus sections are separated by a verse sections i.e. 4 sections containing Verse, Chorus, Verse, Chorus then getting consistent results is much harder as the model doesn't always use the context of the first chorus to generate the second. Perhaps a larger context window like in text models will solve this.
Using auto generated mode, you can actually tell it what to do structure wise. You can ask it to do the same chorus, change it up a bit etc.
Would u mind telling us how ? I tried to get the same Chorus but it's not working ! Always the music changes,lyrics are same !
@wizards-themagicalconcert5048 unfortunately, right now atleast, the only way to get a consistent chorus, is by cramping a verse and a chorus in each 33 second generation. Example: First generation (Verse+Chorus) Extension (Verse 2+Chorus) and so on
@@memesandfun1 Thanks for the info !
I love this channel, so much to learn and it's explained so well. I tried to create a song with the same melody and lyrics but in two languages. Unfortunately, I couldn't find a way to do it. Now, I have two songs with the same lyrics but different melodies. Maybe that will be possible in the future.
Interesting! can't say I've tried that but yeah I suspect it will be difficult to do at this time. It's still very early for this kind of AI but over time I can see more flexibility and consistency.
Very good tips! I've been having issues with the consistency when my songs extend beyond about 3 minutes, and I don't have the chorus appear within the last minute or so, I'm guessing it has to do with the memory it retains. Also ending the song in a similar fashion as it begins is hard.
Glad it was helpful. They did make some updates recently so that getting consistent choruses are better. I believe now it can look back 2min so it has improved memory. Will post an update soon
complimenti
Thanks for the very helpful video. Do you know how I can add a custom thumbnail to my song? Currently it is just the generic red blotch but I see others have ai generated colorful images.
I cover it hereth-cam.com/video/h_Ry2VI01fA/w-d-xo.htmlsi=pn1ZzD21smab7PDb&t=455 👍
Are you Filipino dude 🇵🇭
Thanks so much ! That really helped
You're welcome! Should be better now with the updated context length which is 2min
I'd really love to have the Suno option of extending a song from a specific point. It would make things so much easier! Currently, I sometimes get outros that I like, but that don't come to a proper conclusion, so that I have to add another outro (instrumental) to get the song to actually end - provided, of course, that I don't exceed the 4m22s limit.
I'm sure it will come, considering it's beta it's pretty darn good. I'm excited to see how far they can take this!
when you extend you can activate "crop" and you select what part you want to extend. It's even better that with suno because you can set the beggining point too. With suno you can't extend before... I really hope it will change because that's really limited...
Sometimes I can't choose between the 2 choices (or more if you generate again) that Udio gives you, so I develop both and then remix in a DAW. Had some great results that way and it means you can extend beyond 4 minutes 22 seconds, if your song needs to breathe
Yeah same here but not everyone knows how to use a DAW. With that being said they made some updates including longer songs. Up to 15min if I remember correctly. Will be making an update video soon.
This may be worth mentioning: When I get so far in making my song that I try to remix it, I get a message that it is too far.
Yeah you can only remix the first clip, would be nice to be able to do it in the separate sections.
Who can tell? What to do if you have already typed your text. I like the melody. But you decided to change the text a little. How to edit it while maintaining the old melody?
omg thank you so much for this... I wasted half a day and 200 tries trying to get something consistent and decent. I also challenged myself to write something in Korean xD which I don't recommend (Their system is not suitable for non-english song yet)
It can be stubborn a sometimes and can take quite a few tried to get it right, the model is still very new. I was very curious how it was with other languages. Generally though, the automated lyrics are not the best. Suno AI does a better job, or even chatgpt like I showed in this video 👍
thx for this video. very helpful!
i have seen some ppl use [verse 1] [verse 2] [verse 3] etcpp vs [verse] [verse] [verse]. does this have an impact of the melodys consistancy? Same question goes for a [chorus].
so, does udio add variety if i use differeent numbering?
thx
You're welcome! At this point I can't say it works consistently, I've tried it with numbers and without. What I do know is that it seems the success rate of getting the same verse and chorus is more likely if you put it on the first clip, then extend before or after. I know it's not ideal but as of now that's what I've done to get some consistency. It's still very early so I expect improvements to be made to the model eventually.
My experience after a few hundred hours with the tool is that it gives a through composed melody more often than not.
I’d agree to that. 👍🏼
How do I generate several songs keeping the same voice as the singer?
Not possible at the moment however you could use the same tags and manual mode and get a similar vocal sound. Especially if you put in the style of a particular artist. The downside is it’s totally random and could end up using a lot of credits.
What to do if you want to change a word in the lyrics? same prompt, same sound, just words need to change?
Not possible at the moment unfortunately. Hopefully in the future.
Actually to think of it you can use remix and increase the slider all the way to similar. Although I haven’t tried it myself it could work, curious to test it myself now 👍🏼
How do you prevent repeating verses this is something that is pissing me off in the worst way.
You mean in places where it's not meant to? It does happen every now and then, you can increase the "lyric strength" in advance settings but not too much otherwise it will sound artificial. It's just the nature of the model at the current state. This will improve over time.
@@MonzonMediaLike I write my own lyrics so I may change something and it will keep saying the same verse that I changed before I changed it. Is 60 good for the lyric strength?
Привет! Спасибо за видео. Udio очень классный продукт, я повеселился, когда делал свои песни. Но, как только мне потребовалось сделать что-то сложнее, я потратил порядка 200 генераций, чтобы закончить песню. Приходилось прибегать к различным ухищрениям и редактировать текст таким образом, чтобы AI могла его уместить в размер и правильно расставлять ударения. Чтобы куплеты и припевы звучали единообразно я использовал такие пометки как [verse same as previous verse]. Udio выглядит как перспективная AI и, надеюсь, создатели усовершенствуют её. Сейчас с ней сложновато работать, т.к. это огромнейший рандом, а хотелось бы больше возможностей управления.
can you choose your own tempo ?
Doesn’t hurt to try to put into the tags, some claim it works but from my testing it’s just random. You have better luck using upbeat, slow tempo, aggressive things like that.
I'm trying to follow your instructions for replicating the chorus, but whenever I turn on Manual Mode, I get an error saying my generation attempt failed moderation checks, and no matter what I change, I still get the error.
They were having site issues earlier today. Should be resolved now. Make sure to close down your browser and restart.
@@MonzonMedia Still an issue. It's odd because the lyrics were auto-generated in the first place, and if I turn off manual mode it works with the same lyrics. I guess that's why they call it a beta? 😆Thanks, though!
@@Codester70 interesting working well for me but yeah, considering it’s beta it’s expected.
Man, that is a very powerful tool. Have any idea of how to make money with it? I think making customs songs for birthday parties can be one option. What you guys think?
It’s in beta still you don’t completely own the rights. Likely when the paid plan comes out.
@MonzonMedia A copyright can only be put on a part that a human created/edited. No one can own the part that the AI created. -Top Music Attorney. Check her channel.
Thank you author for your work! But avoid lengthy, unclear expressions, and especially ambiguous ones. You have to re-listen several times to exactly what you meant! Express your thoughts in military terms, this is a program after all! Not a fantasy poem! The FAQ of the PROGRAM is also written in a florid, lengthy and not specific way! We want specifics from you! Good luck!
10:16 the lyrics... oh my... :D
? End of the video?
if you can keep the 2nd verse and the chorus within the 33 seconds on the second clip, then yes u will get the same chorus. but if your 2nd verse fills up the 33 seconds, and u wanna get to the same chorus after that, say goodbye to that chorus. it wont repeat. ive wasted 100s of credits. if anybody know how to get the same chorus again, pls tell me.
p/s: i dunno if it it's better or not, but i generate the chorus first, then build up from there. i always start with chorus, then add before, add intro, then add post-chorus, yada2. then outro.
i find chorus first, then work my way with the verses, etc.
unless u add [intro] [verse] on ur first gen, then maybe it'll generate it properly.
Yes as mentioned it depends on the tempo+lyrics. It’s currently a limitation of the model and I suspect it’s how they preserve the quality without it going off the rails. Sometimes it will take several tries that’s the nature of generative Ai.
wow 😁
Where do my downloads go?
On PC it defaults to your downloads folder.
I just made a track in Udio that is 4:55 duration.
interesting!
Not as impressed with this as I first was, lyrics are clear, but the music all sounds like punk with monotone vocals, clear, crisp, monotone vocals, that being said putting [spoken word] is interesting in an all instrumental
try different genres :D
Remember this is still a beta model and even with Suno AI, it can be very random. Might take a few tries sometimes. It's also free beta at the moment. There is still a lot of experimentation going on, even the developers can't say for sure what works.
I use it to make the lyrics or get an artist...i make beats so i manipulate the hell out of it and im shocked at the results...I just signed my first AI artist to my new imaginary label
i got a longer result by adding intro last. it messed up the filename and wouldent make a video but i could dl audio and cobble this together w yt movie maker. th-cam.com/video/ci27ChXhq78/w-d-xo.html
They've made some updates recently that I will cover soon, now we can create up to 15min songs!