Some additional i use within the lyrics box: [Intro] [scratching] [crowd noise] [cheering] [bass drop] [drum break] [upbeat music] [cheers and applause] [applause] [crowd laughter] [Outro]
@@orwhat24 I know some people, paid plans and on multiple services do the vocals and music separate and mix them after....it seems like the voice always messes my instrumental masterpieces up, yet I have not tried it yet
Not to sure how people give names & access. But, if you want to switch the genders of those vocals you have to add it to the meta tags like: [Chorus: Female Vocals Chorus], [Verse 1: Female Vocals Verses] -yes, the “Vocals” is required. When you set your style/prompt you must declare within the meta tags like (not your lyrics, the prompt: style) like: [Male Vocal Chorus], [Female Vocal Verses], [Male and Female Vocals Duo] (Tip: Correct, odds are 90% of the time the [Male and Female Vocals Duo] won’t make them sing together. It’ll make them blend their voices together. You have to add it as ”sectional” for them to sing together, brother!) Hope this helps!
Just bought this app two days ago…I’m a songwriter and I have been enjoying this app. I have input my songs and love them, but this has put a whole different perspective on how to make them even BETTER!! Thank you!!🙏🏾
I worked with Suno a lot the last couple of months, this has turned out to be it's own kind of science I figured :D It's like a Lion which you have to figure out yourself how to tame. It's wild.
I use "Mariachi", "Spanish Rock", "Reggaeton", "Witch House", "Counterpoint", "Key Change", "Double Bass", "Funny!" in my style prompt to get music that reminds me of El Centro/Mexicali. I also write diphthongs 2-4 in verses leading up to the chorus which has much more and some onomatopoeias. The witch house adds interesting sound effects and oddly pleasing dissonance that I find enchanting. I write "enchantment" in the style sometimes. Mon Laferté inspires me with her writing. I use google translate to translate her songs to English because I always used an offline translater when in Mexico. Ahh the memories!
In the Style of Music box, don't use the enter key after every input. Use commas and spaces. Using all caps in the brackets in the Lyrics box also helps. You have to yell sometimes to get children, and adults for that matter, to do things. Works the same for Suno. It's a child. You must be very specific and detailed or it doesn't get it.
Yes, it is basically a hit or miss with SUNO. I am a paid subscriber, so I upload samples so that SUNO gives me the exact feel that I want. Sometimes, it will follow, but then it will add its own instrument to it😮, or sometimes it's a great surpris❤. But the voice models, that appears to be issue. I have certain genres that I am focused on at this time. One genre is Hip Hop-Soul. I prefer different voice models to match the feel of the song. For example, the music track can be a heavy hitter, with a heavy bass line and drum...SUNO generates a lightweight vocal that does not fit😢. Good luck, everyone. If I come across a new discovery, I will let yall know. Also, I emailed SUNO about voice models today. Let's see if I get a response.🎉
hello mr. dean! first of all, thank you for the great informative video. i've been working for a few months now and i've also been working hard on the subject of “duets”. maybe a little tip, give suno the chance to know in the “music styles” that there should be several voices. the following prompt helped me for the songs “the last match” and “Mailbox”: slow soulful atmospheric ballad, hoarse male voices, high quality sound, violin and in the text field I simply have the instructions [female verse] and [male verse]. Here is an example: [Start] mmhmm [Melodic Built Up] [Intro] mmhmm This match deserves our reverence This stem, our only path from hence (One spark, a chance to burn again) [Bass Drop] [Melodic Interlude] [Male Verse] Your love has grown bitter and cold, (cold) Your stare has left my heart on hold, (hold) Affections faded, passion slowed, Warmth remains in mere aglow! (Warmth remains in mere aglow) [Short Melodic Interlude] best regards from germany
FYI- I do up to 40 “takes” with assorted instructions and tweaking of lyrics before getting something I want. Curiously if I have put the time into the lyrics it sometimes hits a home run on the first two renders. A complaint I have is over instrumental layering as songs go on. I’d prefer a control for that- what NOT to do. Meanwhile, after the loss of dexterity of fingers, I love being able to be musically creative again.
I think the generative AI concept is that it uses a random noise pattern to build up the song. The system then tries to pattern the voice in based on it's neural network, etc. etc. It would be great if within the "noise" pattern it could plant a few [voice_seeds] for the lyrics and we could label those seeds in the lyrics. Like [VOICE: MALE 1] and [CHORUS: CHURCH CHOIR 12 FEMALE], etc. etc. But hey, Suno rocks, bottom line. We're at 3.5 and the music studios are trying to shaft them too. 6 out of 5 stars for their developers and I have no real complaints. I'm sure they are quite dedicated to improving this platform. When 4, 5, and 10 are out these little things will just be things we will laugh at. #Suno
One thing I've found, if you use say guitar balled in the actual lyrics, it does it. If it doesn't want to use the right instruments, do the instrument multiple times in the style. AI needs to be told repeatedly what you want. Think like you're a mad father that cant stop repeating himself. Thats what you need to act like to the AI.
Thanks! I think you’re right, and we’re all just trying to figure out what language it’s speaking so we know what to tell it so that we get what we want. It would be helpful to have a guide - but I guess that’s what we’re all creating now!
@@realmattdean there kind of is but isn't. There's a megathread on the sub reddit, but you kinda have to piece it all together. I also found today that commas work better than periods to make a pause. Like instead of using the elapses... use ,,, instead. The song I just made has alot of tricks I've figured out. I'd link it so you can see what info I gathered, but can't link on youtube.
@@oldmanbiscuit7518 I found using all caps in brackets helps too, like... [INSTRUMENTAL INTRO] or [SPOKEN VERSE] or [BOMB FX]. To repeat a lyric part with background vocals at the end of a line like an echo, I use like... fall apart (fall apart). If you want the vocals to yell, I use all caps and !!! If you include a [BRIDGE] or [INTERLUDE], right under it you can include the instrument to be used like... [HEAVY GUITAR] or [RECORD SCRATCHING]. I'll try the commas instead of being on my period all the time for the pause. Thanks for the tip.
I'm almost certain suno doesn't even know themselves what commands work, lol. Once I wanted it to sound like a boyband and I had to put in "boy band, boyband, boy band, boyband" in there until it finally got it.
I got something that I have not seen anyone use. But I tried it and it seems to work (Most of the time, as anything with Suno) Try something like this: [Strong Vibrato Vocal] | [Pure Soprano Vocal] | [ Slightly Frayed Baritone] | [Clear Vocal] | [Husky Contralto Vocal] | [Nasal Bariton Vocal] | [Light falsetto male] OR - Direct Vocal Ranges: [C4 to F4/G4 Vocal] | [A2 - C5 vocal] | "Disturbed" David Draiman - [E♭1 - F♯5 Vocal] - I had success with these. But I have not tried it enough. Now think the same about Cords? Something like this below here... [Verse Guitar] (G5) (C5) (D5) (Em) or [G5 - C5 - D5 - Em] In a city made of neon light, (G5) (C5) (D5) (Em) or [G5 - C5 - D5 - Em] I roam the streets, feel the electric night. Women's voices are divided into three groups: soprano, mezzo-soprano, and contralto. Men's voices are divided into four groups : countertenor, tenor, baritone, and bass. Then you got the: Ad Lib_ Using Ex. (YEAH!)
I don't understand Suno's obsession with piano. Every time I try to make a relaxing instrumental, despite specifying a particular instrument (harp, pan pipes etc), Suno thinks I want piano with as many notes as possible crammed into every bar.
a tip to make spesific voices: make a suno song that you like, put it through a ai voice changer to the voice that you like, and put the version that you like back into suno, and make a new song out of the voiced changed clip voice changes get lyrics wrong, suno has the text to know what the lyrics are, not only the sound, so if you have the input voice you want, you get a way better result
I hope V4 gives us something like this because sometimes Suno can be hell what works for me every time is this [Verse 1 smooth male voice] i can replace smooth with sultry and more i get 100% male and female if i replace male with female but no control over the voice that it creates. Suno is not perfect yet, jazz fusion, orchestral, sax, electric guitar, new age, progressive, electric bass, smooth male singer, energetic, zen is my prompt used many times i will ask for a guitar solo and a sax solo comes out and a sax solo but guitar solo comes out, so there is a lot of work to be done. I created around 200 songs so far.
Thanks for this video. I am currently transforming some of the lyrics from the Lord Of The Rings books into songs and I was wondering if I could make 2 men sing at once (to make a group of dwarves sing, for example). Don’t know if it’s possible but I’m gonna give it a shot with your tips. Suno has improved a lot over the last six months and I am pretty sure we will be able to fine tune the prompts with much more accuracy and simplicity in the near future. Ps: as you can see, English is not my second language so please forgive me any mistakes! Cheers!
I’d love to hear how it goes! Maybe using the words choir? Or men’s choir? Or men’s duet or quartet? Sounds like an amazing idea. Excited to hear what you find!
I have found that the style of music sometimes influences the vocals. For instance, when prompting "flamenco", the singer will have a raspy voice with a Spanish accent. Same thing with Arabic: Even though the lyrics are in English, the singer will have an Arabic accent (occasionally)
Fantastic Thank you for your help with SUNO. I've been taking notes and I'm going to test it to see how it behaves with Portuguese music from Portugal.
@@realmattdean In the meantime, I'm going to try some tips with SUNO and in the meantime I'll share what happens to me ;) Sorry for my English, but I'm a zero. Eheheh
I've had consistent results from putting vocal prompts in the style box instead of lyrics. For example: Metal, Fast, Nordic Female Vocals. Haven't had any luck with duets yet, but this works very consistently for accents/vocal styles. (Indian, Nordic, British, Latin, etc.)
Unfortunately, we can't ask for many specific things and expect it to work exactly as we would like. I'm certainly going to try the duets, but from what I can see, despite the very good songs we've created, we have to either like them or throw them away. And wait for the technology to develop a little more. But I'm very happy with the results I've got.
I think that you’re right. It seems to me that these songs are build on samples of other music rather than the basics of music theory and the building blocks of basic rhythm and chord structures which make it impossible to direct this ai to make anything other than basic copies of other music. Perhaps in the coming years this will change, but for now - you are correct. We can either accept it for what it is or move on.
@@realmattdean Yeah. But I still remember the first day I was working listen my own playlist made with my own songs.. With the style I like, talking about subjects I like. For People that miss a time when the music talked with him, it's a game changing. 😁. More happy I stayed knowing other people also enjoy the songs I made. Haha. Congrats for the channel! One more subscription here. 😉
@@realmattdean Updating: I made some duests using different prompts like Male Voice/Female voice and others. I created some ozm songs but, it simple stoped to create the duets didn't matter what prompt was.. (It was just one day and I didn't try anymore).
You are doing great working and building an engaging followership with your work around Suno. Well done. But I would like to know if creating an album with the same male or female voice in different tracks and instruments is possible. I will wait for a response from you or anyone who knows how to. Once again, THANKS.
Thanks! I appreciate the kind words. So far, consistency seems to be a really big issue. Unlike ai video which is working hard on character consistency, there is not (yet) any way to prompt the singers to be consistent. Hopefully they will change this soon.
Good luck getting the same voice twice. I have played around with suno and created dozens of tracks. The actual specific voice seems to be a bit random. You can kind of specify mail and female voices, but as far as getting a specific voice repeatedly, good luck.
hello everyone, i think you could use a trick, if you create a beginning where maybe the INTRO is only a mmhhmm or oohhoo and then use the extendend function for example from second 10 or so. so you could manage to get the same voice. the problem would be that the songs start the same, but you could rework that with audacity. best regards from germany
something I do is extending the song, then change the music style to fit the different voice. it's a longer process but it will eventually work haha. I cut the song into three sections
Well I Sampled My Voice and after lots of trys I Got this Great Voice way Diferent than the standard voice that Suno Gives. Waiting to see If my Little Sister comes by to sample her's too.
Well what I'm trying to do is a female sing the song but at the end use five different male voices for each line, so I don't think this will help me but ty thank you for taking the time to make this and you've got a new sub 😉
The duet does not follow the specific lyric prompt but is controlled by the Suno algorithm. I say this because using their own song description as a duet, I get she and he in the lyric, but each line only sometimes follows accurately. Also, if a male and female vocal is part of the style of music area, you get only one of them.
Putting the instrument in the "Style of music" section doesn't work at all but it works pretty well if you put it directly into the lyrics like [epic electric guitar solo]
I believe that some styles obey more than others. It happens to me too, that in certain styles Suno simply ignores the commands and in others it gets closer to what I want.
Assuming that names could work, I'd only expect expect names to work if let's say, you look on Genius for lyrics, and some song examples actually name the singers. But, that assumes that Suno didn't scrub names from the lyrics before putting them in the model. Also, for things like the Latin style... Why not Google Translate your style into some spanish thing?
Quick tip, in the song style box if you put Jamaica-Acent, armenian-accent etc.. you will get what you are looking for. Hope this helps out. Let me know your results.
They really need to find a way to get prompting more accurate and they definitely need to put together voice/band/style ID numbers so we can make consistent sounding music song to song. Right now it's all very random. It's fun, but very random. Everyone losing their minds over the impacts on the music industry need to calm down. It's early. It's random. It still falls into dumb AI holes (prompting as an example) and still will get robotic sounding on some of the best takes.
"Male vocals" or "female vocals" generally work for me, as do style indicators such as grunge or indie. I.e. "male grunge vocals" will generally produce a rough male voice
Just found out about Suno and I'm having a bit of trouble, I want a robotic voice like IBM 7094 but keep getting human-like vocals no matter what I've tried.
I make a lot of rock music I started putting in, stong male voice, in the styles if i don't i may end up with some great music and a very soft male voice
@matt Dean Film if you use Brackets and capslock, inside your lyrics you will or having better chance to get as you want, but you maybe not get it as you on your first create but if you are lucky you get it on 2nd try or 3rd.. that works for be so give that a try
Sometimes, the voices in the songs I'm producing are distorted, sounding like the waves are clipped (overloaded). Do you have any ideas how to fix this? Thanks
Ive made duo's signature style, it works good, but when you put it specific tag on lyrics it may not and may working, it takes me 15 times to prompt but it doesn't perfectly do the job. The 15 times tries was for my trio male rappers and one female melodic vocals. So, if keeps gone wrong, just prompt it again. But I hope Suno AI would fix this, it'd be beneficial for all of us.
@@realmattdean I've made a few songs with Suno that sound suspiciously similar to well-known artists. This one was clearly trained on/a clone of a singer who had a very distinctive voice: th-cam.com/video/vjBC6Sjjvyo/w-d-xo.html
Where do you plan on using these AI voices? Are you planning on submitting these songs with the AI voices to sync licensing companies to be used in commercials? Whats the end goal?
Have you tried (Latin Male Vocals) or whatever style of genre your looking for? Also try putting in the Key you want the vocals and it also works with the music! You'll be surprised by how much change you will get.
Those are great questions that need explored. I wonder how Suno interacts with prompts from other languages with requests like that? May have to make that another one of the videos!
Hi, I got ChatGpt to write in English and translate chorus in Thai, put the lyrics on Suno , th-cam.com/video/VYyWq8I2cWA/w-d-xo.htmlsi=84SQbN6_nRuS91tX
Hi, I got ChatGpt to write in English and translate chorus in Thai, put the lyrics on Suno , th-cam.com/video/VYyWq8I2cWA/w-d-xo.htmlsi=84SQbN6_nRuS91tX
You can do spoken words using the prompt [Spoken Dialogue] before the text.. Don't use quotation marks, or brakets for the words after prompt [Spoken Dialogue]. as far as singing without music you could try using something like [Sung Dialogue] or [Singing Dialogue] & or also try prompt [No music] followed by [Acapella vocal]. No promises, but that could, and should get results so give it a go. Good luck!
Thanks for the video, even though I am working with suno for the last one month and tried to create a duet song and after experiment I am able to create a song which is very close to what I want, as links we can't share , if you wish let me know so I can share the link , your videos are really helpful for people like us who are new to Suno Ai. Or alternatively you can search Bollywood Ai_music and the song is wada karo. Please let me know your valuable comments.
I just joined Suno, and found it interesting how it takes some lyrics and writes the whole song. It would be nice if you could give it a basic music line of the song above the lyrics, including the chorus, and force it to stick to your original idea because the songs are nice, but it's not how my original soundtrack was. I am new to this whole world of IA, so I am struggling a bit.
I'm new to this..literally just started..I found music I liked but felt some of my lyrics needed to change to suit the music more..how do I do that without having to regenerate and waste credits ? I'm using the free app atm
@@realmattdean Thanks! Do you mean generally or regarding duets? Anyway, I think the reason it's so hard to make a duet in Suno is partly because it's not an official feature and partly because Suno seems to regard 'male' and 'female' in the commands within song lyrics as simply labels rather than knowing what they mean in the way that it does with them in the style description.
@@themachine5957 Generally. feel free to share. Ya - I keep hoping that we can get more descriptive prompts within the song to be a better guideline for the song creation. It will SOMETIMES do what you ask and other times do what ever it wants. seems pretty obvious that it's following some sort of predetermined path rather than truly following the prompts.
Si se ponen a pensar... La IA se programa como si se estuviera hablando... en realidad le estas hablando a un "Niño" el cual no entiende quien es [Mujer] y [Hombre] la IA al tu escribir eso creara diferentes voces. En cambio si ponemos [Mujer1] y [Hombre1] mantendra su voz. Es un sistema complejo pero la IA tiene edad. Se tendria que abrir una escuela que enseñe la "Filosofia de la robotica" y la foroma de pensar "Pensamiento de la robotica 1".... En fin me interesa mucho el tema pero en argentina no tenemos nada para estudiar algo asi.
Had a lot of fun with Suno. But recently, despite the updates, It seems to have become harder to get it to generate something close to what I'm after. Vocals are the biggest issue for me - they're all starting to sound increasingly similar, depending on your chosen style prompts. I can't help wondering if they are getting nervous about lawsuits for voice cloning. I made this one a while back and it's clearly a copy of a well-known artist: th-cam.com/video/vjBC6Sjjvyo/w-d-xo.html
I myself have tried almost everything possible to change the voice of the sungers. All the songs start to sound the same when they are always sung by the same type of singer. For blues and soul, I would like the singers to have much more rawness.
I agree and I’ve run into the same problem.. I’m going to try more descriptive prompts regarding the singer in the description box and see if that helps. Let me know if you find anything that works.
it is very bad on duet stuff.. sometimes it randomly does great job.. mostly doesnt.. i hope they add those feaures in the next update.. at least we must choose the gender of voice male or female on the settings without mentioning at prompts or lyrics
The only thing I have found that really remotely even works is to kind of specify Spanish voice or soulful voice etc. I have had little to no luck specifying specific instruments for songs or even parts of songs. That all seems to be kind of random so far.
At first, Suno seemed promising with its free version, but once I paid for it, I started getting frustrated. No matter the genre, the AI kept using the same types of instruments and voices, which made the music feel repetitive. What irritated me even more was hearing other users' songs and realizing their artists either sounded just like mine or had that unmistakable "Suno" sound. It became obvious that the platform lacked variety compared to others.
Using names like Marvin and Aretha and hoping it'll sound like the actual singers is pointless. Suno trains their models on data with all mentions of any artist names explicitly removed, to prevent people from asking it to copy real (and copyrighted) music. So the AI literally has no idea who any artists are; all the songs it's learned from are, in its understanding, anonymous. I don't know how much this matters, but try separating your style tags with commas instead of line breaks. I haven't seen anyone use line breaks, and I get the feeling the model was trained with comma-separated tags. At the very least, it seems to just remove the line breaks, and I'd expect "guitar ballad, slow" to be clearer to the model than "guitar ballad slow" (with the latter being more likely to consider "guitar" and "ballad" separate and equal descriptors, rather than one unified and related descriptor).
Sounds like some good ideas. I’ve experimented in other videos and found that it does seem to recognize basic music language. There are quite a few prompts involving tempo, and specific instrumentation that it seems to recognize. It would be helpful if they were a little more clear on its capabilities and functionality.
@@realmattdean Thing is, I'm not sure they know. They collected and preprocessed the training data almost certainly automatically, then trained it up. It's a Transformer based model, meaning it's self-supervised, so what it learned from the training data is more or less a black box. All they can do is try stuff and see what it does and doesn't understand. The Suno Discord has a "technical discussion" channel where people often help with techniques for improving prompts, and people there have discovered a ton of tags Suno understands, if you wanted to learn more.
I tried until today create songs and it looks fine but today it start make weird things.. first cover image in arabic style then song itself was like prayer for muslim people.... when I type Musical What is Islamic on Musical? Hmm i wonder, then other next try of that song came with weeeird male haard soo deep voice with raspy voice and too quick spelling.. I dont know what happened
As a Brazilian user, I can say that in rhythms here in the country that are already available, the voices are still poor and the commands do not respond well. Rhythms and commands in English, or even Asian ones, respond much better.
![ตีสิบเดย์ [FULL] | เปิดใจ "อาจารย์ เบียร์ ฅนตื่นธรรม" นักสนทนาธรรมสายดุดัน!](http://i.ytimg.com/vi/vqaBI_mnoDk/mqdefault.jpg)
Some additional i use within the lyrics box:
[Intro]
[scratching]
[crowd noise]
[cheering]
[bass drop]
[drum break]
[upbeat music]
[cheers and applause]
[applause]
[crowd laughter]
[Outro]
They should just have a save voice feature and a remix voice feature for using them later.
agreed
Super idea.
@@orwhat24 I know some people, paid plans and on multiple services do the vocals and music separate and mix them after....it seems like the voice always messes my instrumental masterpieces up, yet I have not tried it yet
That's the ticket, Teleporter!
Yesss!!!! But it’s easier if you find a tone like “cheerful” & kinda design your own “tone”. A cool little workaround for ya. 😉
Not to sure how people give names & access. But, if you want to switch the genders of those vocals you have to add it to the meta tags like: [Chorus: Female Vocals Chorus], [Verse 1: Female Vocals Verses] -yes, the “Vocals” is required.
When you set your style/prompt you must declare within the meta tags like (not your lyrics, the prompt: style) like:
[Male Vocal Chorus], [Female Vocal Verses], [Male and Female Vocals Duo]
(Tip: Correct, odds are 90% of the time the [Male and Female Vocals Duo] won’t make them sing together. It’ll make them blend their voices together. You have to add it as ”sectional” for them to sing together, brother!)
Hope this helps!
Just bought this app two days ago…I’m a songwriter and I have been enjoying this app. I have input my songs and love them, but this has put a whole different perspective on how to make them even BETTER!! Thank you!!🙏🏾
I worked with Suno a lot the last couple of months, this has turned out to be it's own kind of science I figured :D It's like a Lion which you have to figure out yourself how to tame. It's wild.
I agree, do you know of any Suno groups that converse on techniques they use "tame" it?
I use "Mariachi", "Spanish Rock", "Reggaeton", "Witch House", "Counterpoint", "Key Change", "Double Bass", "Funny!" in my style prompt to get music that reminds me of El Centro/Mexicali. I also write diphthongs 2-4 in verses leading up to the chorus which has much more and some onomatopoeias. The witch house adds interesting sound effects and oddly pleasing dissonance that I find enchanting. I write "enchantment" in the style sometimes. Mon Laferté inspires me with her writing. I use google translate to translate her songs to English because I always used an offline translater when in Mexico. Ahh the memories!
Pop choir gets my duets working more than other genres.
In the Style of Music box, don't use the enter key after every input. Use commas and spaces. Using all caps in the brackets in the Lyrics box also helps. You have to yell sometimes to get children, and adults for that matter, to do things. Works the same for Suno. It's a child. You must be very specific and detailed or it doesn't get it.
Yes, it is basically a hit or miss with SUNO. I am a paid subscriber, so I upload samples so that SUNO gives me the exact feel that I want. Sometimes, it will follow, but then it will add its own instrument to it😮, or sometimes it's a great surpris❤. But the voice models, that appears to be issue. I have certain genres that I am focused on at this time. One genre is Hip Hop-Soul. I prefer different voice models to match the feel of the song. For example, the music track can be a heavy hitter, with a heavy bass line and drum...SUNO generates a lightweight vocal that does not fit😢. Good luck, everyone. If I come across a new discovery, I will let yall know. Also, I emailed SUNO about voice models today. Let's see if I get a response.🎉
You have opened a new universe for me with this brackets parametering ! THX!
So glad it’s been helpful!!!
hello mr. dean! first of all, thank you for the great informative video.
i've been working for a few months now and i've also been working hard on the subject of “duets”.
maybe a little tip, give suno the chance to know in the “music styles” that there should be several voices. the following prompt helped me for the songs “the last match” and “Mailbox”:
slow soulful atmospheric ballad, hoarse male voices, high quality sound, violin
and in the text field I simply have the instructions [female verse] and [male verse]. Here is an example:
[Start]
mmhmm
[Melodic Built Up]
[Intro]
mmhmm
This match deserves our reverence
This stem, our only path from hence
(One spark, a chance to burn again)
[Bass Drop]
[Melodic Interlude]
[Male Verse]
Your love has grown bitter and cold, (cold)
Your stare has left my heart on hold, (hold)
Affections faded, passion slowed,
Warmth remains in mere aglow!
(Warmth remains in mere aglow)
[Short Melodic Interlude]
best regards from germany
So far I have enjoyed all your video's on Suno.
Thank you! I hope they’ve been helpful!
This is wonderful. I really liked the tests. I also do tests, but seeing who understands more opens your mind to try new things. Thank you!
FYI- I do up to 40 “takes” with assorted instructions and tweaking of lyrics before getting something I want.
Curiously if I have put the time into the lyrics it sometimes hits a home run on the first two renders.
A complaint I have is over instrumental layering as songs go on. I’d prefer a control for that- what NOT to do.
Meanwhile, after the loss of dexterity of fingers, I love being able to be musically creative again.
If you can, use Extend and Covers to make songs longer and broader. Cause you change the Beat in the covers as an Example.
Sometimes I think it does that to drain credits. 😂
I think the generative AI concept is that it uses a random noise pattern to build up the song. The system then tries to pattern the voice in based on it's neural network, etc. etc. It would be great if within the "noise" pattern it could plant a few [voice_seeds] for the lyrics and we could label those seeds in the lyrics. Like [VOICE: MALE 1] and [CHORUS: CHURCH CHOIR 12 FEMALE], etc. etc.
But hey, Suno rocks, bottom line. We're at 3.5 and the music studios are trying to shaft them too. 6 out of 5 stars for their developers and I have no real complaints. I'm sure they are quite dedicated to improving this platform. When 4, 5, and 10 are out these little things will just be things we will laugh at. #Suno
One thing I've found, if you use say guitar balled in the actual lyrics, it does it. If it doesn't want to use the right instruments, do the instrument multiple times in the style. AI needs to be told repeatedly what you want. Think like you're a mad father that cant stop repeating himself. Thats what you need to act like to the AI.
Thanks! I think you’re right, and we’re all just trying to figure out what language it’s speaking so we know what to tell it so that we get what we want. It would be helpful to have a guide - but I guess that’s what we’re all creating now!
@@realmattdean there kind of is but isn't. There's a megathread on the sub reddit, but you kinda have to piece it all together. I also found today that commas work better than periods to make a pause. Like instead of using the elapses... use ,,, instead. The song I just made has alot of tricks I've figured out. I'd link it so you can see what info I gathered, but can't link on youtube.
Ellipses and use that to trail off with phrases instead of pausing.
@@oldmanbiscuit7518 I found using all caps in brackets helps too, like... [INSTRUMENTAL INTRO] or [SPOKEN VERSE] or [BOMB FX]. To repeat a lyric part with background vocals at the end of a line like an echo, I use like... fall apart (fall apart). If you want the vocals to yell, I use all caps and !!! If you include a [BRIDGE] or [INTERLUDE], right under it you can include the instrument to be used like... [HEAVY GUITAR] or [RECORD SCRATCHING]. I'll try the commas instead of being on my period all the time for the pause. Thanks for the tip.
I'm almost certain suno doesn't even know themselves what commands work, lol. Once I wanted it to sound like a boyband and I had to put in "boy band, boyband, boy band, boyband" in there until it finally got it.
I got something that I have not seen anyone use. But I tried it and it seems to work (Most of the time, as anything with Suno)
Try something like this: [Strong Vibrato Vocal] | [Pure Soprano Vocal] | [ Slightly Frayed Baritone] | [Clear Vocal] | [Husky Contralto Vocal] | [Nasal Bariton Vocal] | [Light falsetto male]
OR
- Direct Vocal Ranges: [C4 to F4/G4 Vocal] | [A2 - C5 vocal] | "Disturbed" David Draiman - [E♭1 - F♯5 Vocal] - I had success with these. But I have not tried it enough. Now think the same about Cords? Something like this below here...
[Verse Guitar]
(G5) (C5) (D5) (Em) or [G5 - C5 - D5 - Em]
In a city made of neon light,
(G5) (C5) (D5) (Em) or [G5 - C5 - D5 - Em]
I roam the streets, feel the electric night.
Women's voices are divided into three groups: soprano, mezzo-soprano, and contralto.
Men's voices are divided into four groups : countertenor, tenor, baritone, and bass.
Then you got the:
Ad Lib_
Using Ex. (YEAH!)
I don't understand Suno's obsession with piano. Every time I try to make a relaxing instrumental, despite specifying a particular instrument (harp, pan pipes etc), Suno thinks I want piano with as many notes as possible crammed into every bar.
a tip to make spesific voices: make a suno song that you like, put it through a ai voice changer to the voice that you like, and put the version that you like back into suno, and make a new song out of the voiced changed clip
voice changes get lyrics wrong, suno has the text to know what the lyrics are, not only the sound, so if you have the input voice you want, you get a way better result
Thanks for your video, I use Suno for a long time,end i found that Subo do always what Suno wants, not what you want not follow your instructions
Thank you sir. I have enjoyed watching this video.I will use this as a guide to make my new song by using suno.
Glad to help!! It’s a lot of fun to use this and make discoveries. Let me know how it goes!
We defo need the ability to point Suno to a Specific voice just as you have done. Suno needs work!
I hope V4 gives us something like this because sometimes Suno can be hell what works for me every time is this [Verse 1 smooth male voice] i can replace smooth with sultry and more i get 100% male and female if i replace male with female but no control over the voice that it creates.
Suno is not perfect yet, jazz fusion, orchestral, sax, electric guitar, new age, progressive, electric bass, smooth male singer, energetic, zen is my prompt used many times i will ask for a guitar solo and a sax solo comes out and a sax solo but guitar solo comes out, so there is a lot of work to be done. I created around 200 songs so far.
Thanks for this video. I am currently transforming some of the lyrics from the Lord Of The Rings books into songs and I was wondering if I could make 2 men sing at once (to make a group of dwarves sing, for example). Don’t know if it’s possible but I’m gonna give it a shot with your tips. Suno has improved a lot over the last six months and I am pretty sure we will be able to fine tune the prompts with much more accuracy and simplicity in the near future. Ps: as you can see, English is not my second language so please forgive me any mistakes! Cheers!
I’d love to hear how it goes! Maybe using the words choir? Or men’s choir? Or men’s duet or quartet? Sounds like an amazing idea. Excited to hear what you find!
gregorian chant might be useful
LOVE this video thanks for sharing, i need to learn from you more and more sir.
I have found that the style of music sometimes influences the vocals. For instance, when prompting "flamenco", the singer will have a raspy voice with a Spanish accent. Same thing with Arabic: Even though the lyrics are in English, the singer will have an Arabic accent (occasionally)
Writing lyrics in different languages also drastically affects the AI's bias
Fantastic
Thank you for your help with SUNO. I've been taking notes and I'm going to test it to see how it behaves with Portuguese music from Portugal.
I would love to hear about what you find! I’m hoping we can compile a list of what actually works with each music style and language. Thank you!!
@@realmattdean In the meantime, I'm going to try some tips with SUNO and in the meantime I'll share what happens to me ;)
Sorry for my English, but I'm a zero. Eheheh
I've had consistent results from putting vocal prompts in the style box instead of lyrics.
For example:
Metal, Fast, Nordic Female Vocals.
Haven't had any luck with duets yet, but this works very consistently for accents/vocal styles. (Indian, Nordic, British, Latin, etc.)
Unfortunately, we can't ask for many specific things and expect it to work exactly as we would like. I'm certainly going to try the duets, but from what I can see, despite the very good songs we've created, we have to either like them or throw them away. And wait for the technology to develop a little more. But I'm very happy with the results I've got.
I think that you’re right. It seems to me that these songs are build on samples of other music rather than the basics of music theory and the building blocks of basic rhythm and chord structures which make it impossible to direct this ai to make anything other than basic copies of other music. Perhaps in the coming years this will change, but for now - you are correct. We can either accept it for what it is or move on.
@@realmattdean Yeah. But I still remember the first day I was working listen my own playlist made with my own songs.. With the style I like, talking about subjects I like. For People that miss a time when the music talked with him, it's a game changing. 😁. More happy I stayed knowing other people also enjoy the songs I made. Haha. Congrats for the channel! One more subscription here. 😉
@@almirsantosmusic thanks! it is fun having a new tool to play with. Looking forward to seeing where it goes!
@@realmattdean Updating: I made some duests using different prompts like Male Voice/Female voice and others. I created some ozm songs but, it simple stoped to create the duets didn't matter what prompt was.. (It was just one day and I didn't try anymore).
im HIspanic and this will give you that mexican music flavor that I use myself. (Regional mexicano acústico tradicional)
You are doing great working and building an engaging followership with your work around Suno. Well done. But I would like to know if creating an album with the same male or female voice in different tracks and instruments is possible. I will wait for a response from you or anyone who knows how to. Once again, THANKS.
Thanks! I appreciate the kind words. So far, consistency seems to be a really big issue. Unlike ai video which is working hard on character consistency, there is not (yet) any way to prompt the singers to be consistent. Hopefully they will change this soon.
Good luck getting the same voice twice. I have played around with suno and created dozens of tracks. The actual specific voice seems to be a bit random. You can kind of specify mail and female voices, but as far as getting a specific voice repeatedly, good luck.
hello everyone, i think you could use a trick, if you create a beginning where maybe the INTRO is only a mmhhmm or oohhoo and then use the extendend function for example from second 10 or so. so you could manage to get the same voice. the problem would be that the songs start the same, but you could rework that with audacity. best regards from germany
I have seen voices Suno come up with depend a lot on the genre.
Agreed. It's because of what ever they are sampling to create that genre.
something I do is extending the song, then change the music style to fit the different voice. it's a longer process but it will eventually work haha. I cut the song into three sections
Well I Sampled My Voice and after lots of trys I Got this Great Voice way Diferent than the standard voice that Suno Gives. Waiting to see If my Little Sister comes by to sample her's too.
By The way it Pitch corrects if you are not that much of a Singer, Did I Say me ? Yeah, ha ha ha!!
I have gotten good results by specifying the singer's country and gender. e.g. Jamaican male voice
Well what I'm trying to do is a female sing the song but at the end use five different male voices for each line, so I don't think this will help me but ty thank you for taking the time to make this and you've got a new sub 😉
I a addicted to suno now.. great experience
Muddy Southern Rap, Rough Male Vocals
Helpful experiments. Thanks
The duet does not follow the specific lyric prompt but is controlled by the Suno algorithm. I say this because using their own song description as a duet, I get she and he in the lyric, but each line only sometimes follows accurately. Also, if a male and female vocal is part of the style of music area, you get only one of them.
[Male Voice] [Female Voice] - [Female Vocalist and Male Vocalist] I looked up duet in the search
Putting the instrument in the "Style of music" section doesn't work at all but it works pretty well if you put it directly into the lyrics like [epic electric guitar solo]
I believe that some styles obey more than others. It happens to me too, that in certain styles Suno simply ignores the commands and in others it gets closer to what I want.
But how can I use the same voice again and again?
Good Explanation, thanks for this video
Assuming that names could work, I'd only expect expect names to work if let's say, you look on Genius for lyrics, and some song examples actually name the singers.
But, that assumes that Suno didn't scrub names from the lyrics before putting them in the model.
Also, for things like the Latin style... Why not Google Translate your style into some spanish thing?
Quick tip, in the song style box if you put Jamaica-Acent, armenian-accent etc.. you will get what you are looking for. Hope this helps out. Let me know your results.
They really need to find a way to get prompting more accurate and they definitely need to put together voice/band/style ID numbers so we can make consistent sounding music song to song. Right now it's all very random. It's fun, but very random.
Everyone losing their minds over the impacts on the music industry need to calm down. It's early. It's random. It still falls into dumb AI holes (prompting as an example) and still will get robotic sounding on some of the best takes.
being able to edit the songs would be great too.
@@peristiloperis7789 yesss ... that's a good thought. After it's spun ... I've had songs that spun so good but had a goofy typo ...
"Male vocals" or "female vocals" generally work for me, as do style indicators such as grunge or indie. I.e. "male grunge vocals" will generally produce a rough male voice
Aggressive also works
Using things like sexy voice gives you a breathier singer
thanks for the video.
keep music volume low while you speak. sometimes it is not clearly audible what you are saying because music sound is high
Thanks - yes, I had the setting wrong when it was recorded and fixed it for the next one. 😊
@@realmattdean And you can also use a sidechain compressor with a long release to automatically lower the volume of the song. =]
That's awesome!
Just found out about Suno and I'm having a bit of trouble, I want a robotic voice like IBM 7094 but keep getting human-like vocals no matter what I've tried.
I feel conflicted about if this is even making music but I keep making songs 😅
I make a lot of rock music I started putting in, stong male voice, in the styles if i don't i may end up with some great music and a very soft male voice
@matt Dean Film if you use Brackets and capslock, inside your lyrics you will or having better chance to get as you want, but you maybe not get it as you on your first create but if you are lucky you get it on 2nd try or 3rd.. that works for be so give that a try
Hi brother, When creating songs on suno Ai or any other Ai sites and publishing them on TH-cam, your opinion about the new Ai altered.
Sometimes, the voices in the songs I'm producing are distorted, sounding like the waves are clipped (overloaded). Do you have any ideas how to fix this? Thanks
Ive made duo's signature style, it works good, but when you put it specific tag on lyrics it may not and may working, it takes me 15 times to prompt but it doesn't perfectly do the job. The 15 times tries was for my trio male rappers and one female melodic vocals. So, if keeps gone wrong, just prompt it again. But I hope Suno AI would fix this, it'd be beneficial for all of us.
is there a way to have new songs done with dead singers?
thanks for the great videos
No - they aren’t allowed to use the voice of any real person.
@@realmattdean I've made a few songs with Suno that sound suspiciously similar to well-known artists. This one was clearly trained on/a clone of a singer who had a very distinctive voice: th-cam.com/video/vjBC6Sjjvyo/w-d-xo.html
Where do you plan on using these AI voices? Are you planning on submitting these songs with the AI voices to sync licensing companies to be used in commercials? Whats the end goal?
i'm trying to get alt-baritone/baritone vocals like till lindeman, doesn't want to work so far.
Have you tried (Latin Male Vocals) or whatever style of genre your looking for? Also try putting in the Key you want the vocals and it also works with the music! You'll be surprised by how much change you will get.
What unique instruments, like the Khaen, Khong Wong Lek OR uncommon languages, like Lao?
Those are great questions that need explored. I wonder how Suno interacts with prompts from other languages with requests like that? May have to make that another one of the videos!
Hi, I got ChatGpt to write in English and translate chorus in Thai, put the lyrics on Suno , th-cam.com/video/VYyWq8I2cWA/w-d-xo.htmlsi=84SQbN6_nRuS91tX
Hi, I got ChatGpt to write in English and translate chorus in Thai, put the lyrics on Suno , th-cam.com/video/VYyWq8I2cWA/w-d-xo.htmlsi=84SQbN6_nRuS91tX
Anyone managed to create with Suno just the vocals (acapella) without the instrumental?
You can do spoken words using the prompt [Spoken Dialogue] before the text.. Don't use quotation marks, or brakets for the words after prompt [Spoken Dialogue]. as far as singing without music you could try using something like [Sung Dialogue] or [Singing Dialogue] & or also try prompt [No music] followed by [Acapella vocal]. No promises, but that could, and should get results so give it a go. Good luck!
Thanks for the video, even though I am working with suno for the last one month and tried to create a duet song and after experiment I am able to create a song which is very close to what I want, as links we can't share , if you wish let me know so I can share the link , your videos are really helpful for people like us who are new to Suno Ai. Or alternatively you can search Bollywood Ai_music and the song is wada karo. Please let me know your valuable comments.
I just joined Suno, and found it interesting how it takes some lyrics and writes the whole song. It would be nice if you could give it a basic music line of the song above the lyrics, including the chorus, and force it to stick to your original idea because the songs are nice, but it's not how my original soundtrack was. I am new to this whole world of IA, so I am struggling a bit.
Good Stuff!
Isn’t this a function of what Suno should be providing, without us having to do a hunt and peck to get out of the tool, the desired outcome?
I need... more credits.
How can I create only acapella?
How come it doesn't look like this on my App?
How do you do multiple voices in same song
I'm new to this..literally just started..I found music I liked but felt some of my lyrics needed to change to suit the music more..how do I do that without having to regenerate and waste credits ? I'm using the free app atm
I, too only started 3 days ago. I'm not sure if this will work but you could try Reuse Prompts.
How to get the instrumental out of the song?
Thanks 🎉😊
Hello can you tell me how can I remove the voice of the singer in the uploaded song, I don't need the voice, I just need the music of the song
Bandlab splitter is free
Yes - you’ll have to use separate software for now to remove a voice from a song you’re uploading.
Well, I wish I'd seen this video a month ago (yes I know it wan't uploaded yet). It would have saved me a lot of credits. 😊
I’m glad I can be helpful now!! Best of luck. Would love to hear what you create.
@@realmattdean Thanks! Do you mean generally or regarding duets?
Anyway, I think the reason it's so hard to make a duet in Suno is partly because it's not an official feature and partly because Suno seems to regard 'male' and 'female' in the commands within song lyrics as simply labels rather than knowing what they mean in the way that it does with them in the style description.
@@themachine5957 Generally. feel free to share. Ya - I keep hoping that we can get more descriptive prompts within the song to be a better guideline for the song creation. It will SOMETIMES do what you ask and other times do what ever it wants. seems pretty obvious that it's following some sort of predetermined path rather than truly following the prompts.
could I download midi? so that I may redo ay song?
try to put "MALE VOCAL" in style of music filed.
Obrigado pelas dicas, abraço!
Pls help me to download my song, i cannot download my song
hi sir how to add my lyrics in audio ?
Si se ponen a pensar... La IA se programa como si se estuviera hablando... en realidad le estas hablando a un "Niño" el cual no entiende quien es [Mujer] y [Hombre] la IA al tu escribir eso creara diferentes voces.
En cambio si ponemos [Mujer1] y [Hombre1] mantendra su voz. Es un sistema complejo pero la IA tiene edad.
Se tendria que abrir una escuela que enseñe la "Filosofia de la robotica" y la foroma de pensar "Pensamiento de la robotica 1"....
En fin me interesa mucho el tema pero en argentina no tenemos nada para estudiar algo asi.
BZRP #56 is one of my favorite songs!
Thanks!
Had a lot of fun with Suno. But recently, despite the updates, It seems to have become harder to get it to generate something close to what I'm after. Vocals are the biggest issue for me - they're all starting to sound increasingly similar, depending on your chosen style prompts. I can't help wondering if they are getting nervous about lawsuits for voice cloning. I made this one a while back and it's clearly a copy of a well-known artist: th-cam.com/video/vjBC6Sjjvyo/w-d-xo.html
[Verse 1 Male]
[Verse 2 Female]
it will work?
try putting tempo. keys for voice in music type fiels and male and femal clear voice / vocals, or laina woman, latino man etc
I myself have tried almost everything possible to change the voice of the sungers. All the songs start to sound the same when they are always sung by the same type of singer. For blues and soul, I would like the singers to have much more rawness.
I agree and I’ve run into the same problem.. I’m going to try more descriptive prompts regarding the singer in the description box and see if that helps. Let me know if you find anything that works.
it is very bad on duet stuff.. sometimes it randomly does great job.. mostly doesnt..
i hope they add those feaures in the next update.. at least we must choose the gender of voice male or female on the settings without mentioning at prompts or lyrics
The only thing I have found that really remotely even works is to kind of specify Spanish voice or soulful voice etc. I have had little to no luck specifying specific instruments for songs or even parts of songs. That all seems to be kind of random so far.
At first, Suno seemed promising with its free version, but once I paid for it, I started getting frustrated. No matter the genre, the AI kept using the same types of instruments and voices, which made the music feel repetitive. What irritated me even more was hearing other users' songs and realizing their artists either sounded just like mine or had that unmistakable "Suno" sound. It became obvious that the platform lacked variety compared to others.
Using names like Marvin and Aretha and hoping it'll sound like the actual singers is pointless. Suno trains their models on data with all mentions of any artist names explicitly removed, to prevent people from asking it to copy real (and copyrighted) music. So the AI literally has no idea who any artists are; all the songs it's learned from are, in its understanding, anonymous.
I don't know how much this matters, but try separating your style tags with commas instead of line breaks. I haven't seen anyone use line breaks, and I get the feeling the model was trained with comma-separated tags. At the very least, it seems to just remove the line breaks, and I'd expect "guitar ballad, slow" to be clearer to the model than "guitar ballad slow" (with the latter being more likely to consider "guitar" and "ballad" separate and equal descriptors, rather than one unified and related descriptor).
Sounds like some good ideas. I’ve experimented in other videos and found that it does seem to recognize basic music language. There are quite a few prompts involving tempo, and specific instrumentation that it seems to recognize. It would be helpful if they were a little more clear on its capabilities and functionality.
@@realmattdean Thing is, I'm not sure they know. They collected and preprocessed the training data almost certainly automatically, then trained it up. It's a Transformer based model, meaning it's self-supervised, so what it learned from the training data is more or less a black box. All they can do is try stuff and see what it does and doesn't understand. The Suno Discord has a "technical discussion" channel where people often help with techniques for improving prompts, and people there have discovered a ton of tags Suno understands, if you wanted to learn more.
17 minuti per dire che Suno ignora i nomi propri e di utilizzare invece [male] [female] and [male and female duet]
Try getting it to sing with a Boston accent.
Love it. Will put that on the list.
use "y-o-o-o-ou" instead of "yooou"
(Guitar ballad )
I tried until today create songs and it looks fine but today it start make weird things.. first cover image in arabic style then song itself was like prayer for muslim people.... when I type Musical What is Islamic on Musical? Hmm i wonder, then other next try of that song came with weeeird male haard soo deep voice with raspy voice and too quick spelling.. I dont know what happened
One suggestion is to use words that indicate the singer's gender... well, it's just another idea.
Try vocal tags instead. Also 3.5 is generic. Try 3.0 for more broad and interesting results.
As a Brazilian user, I can say that in rhythms here in the country that are already available, the voices are still poor and the commands do not respond well. Rhythms and commands in English, or even Asian ones, respond much better.
Como consigo criar um brega no estilo de Reginaldo Rossi?
Just accept the fact that you have little control and stop wasting your money. Just take what it gives you.