I ain't posting my actual good songs since I upload them to YT, don't want to doxx myself, but here's one that I intended to use in a sovietwave WEG I wanted to make during the week I was obsessed with RPGM.

>>9975 Ye I'm pretty much in the same boat here, to a point where I already recognized one of the anons that posted earlier.

Is there any equivalent of Stable Diffusion for ai music? Suno is okey but It feels limited if you don't pay for a subscription.

>>6158 I love these and I love you, anon

>>9975 >sovietwave WEG Based, Very good song. I like the synth, sounds nostalgic. >>12962 >Is there any equivalent of Stable Diffusion for ai music? I dont think so. Sadly open source audio generation is quite neglected, Although theres some alternatives like Stable Open Audio Zero (1, loud warning), Monster Piano ( 2 ), Unlimited Music Gen ( 3, Boring), DiffRhythm ( 4, first 30s are good, the rest is slop), These are the best ones i know (Quite trashy, yeah i know) I guess we have to wait till a good model its released. Maybe a ///chinese\\\ one? lel >>13016 >I love these and I love you, anon Really? kek, those are old trash lol.(No offense and thanks for the gratitude though) I generated those a year ago. The current udio model is way better. Although i got tired of it being propietary and being censured. Nevertheless theres a lot of good gens on udio. For example i found this account that makes really good songs. https://www.udio.com/creators/lunarisNyx24 See (5) I promise is good music this time kek

Yo había realizado algunas con suno y solo por los kekes.

>>13087 >no open source for audio gen Sad. Sometimes I want to know how to compose , play and produce a piece of music by myself just to decompose and reassemble what Suno gives me because that site like to blueball you giving you a nice piece of music for the first 20 seconds and then proceeds to repeat it as a loop for 4 minutes or do something different than you don't like at all.

>>13147 Dont speak spanish but thanks for the songs. The last one sounds very good. I dont usually listen to metal but that one was a banger :^) >>13188 >>no open source for audio gen Yeah due to the strict copyright laws and small databases, Theres not main models yet, but theres some small shitty models like DiffRhythm or YuE (Both suck kek). I still gonna continue searching for something interesting though. >giving you a nice piece of music for the first 20 seconds and then proceeds to repeat it as a loop for 4 minutes or do something different than you don't like at all. Yeah that happens because these diffusion models work in "batches" of 20-30 secs. So later it joins together these parts to make a full song. Sometimes it will fail completely and enter in a infinite loop or hallucinate stuff up. Nevertheless someone will eventually make a good model, in the meantime the only thing we can do is to listen some comfy music kek (Made by LunarisNyx24). Also interesting song, what are the lyrics saying?

>>13245 the lyrics we're writted in english first but Suno can't give me a decent Vocaloid voice if is not writted in Japanese so I used chatGPT to translate it. [Verse 1] Romaji: Mata nebou, yume no naka no hou ga mashi de Makura dake ga wakatteru boku no subete Midoku no mama, sawagashii nounai Bagu mitai na egao de, kyomu wo saketai Toosuto kuwaete, chikoku kakutei Shikou mo teishi, atama mainasu san-paasento Sumaho wa denchi ichi-paasento Boku no shikou, urusasugiru yo ne Translation: Overslept again—dreams were better than real Only my pillow knows how I actually feel Left unread, my mind’s a crowded street Smiling like a glitch just to dodge defeat Toast in my mouth, yep, I’m late again Brain’s gone blank—down to minus ten Phone’s at 1%, mood’s just noise My thoughts are screaming louder than my voice [Instrumental Break] [Pre-Chorus] Romaji: Sekai wa guchagucha, demo kakuseba mashi Mada ikiteru, sore dake de... (Kakusu) (kakusu)... (ikiteru) Translation: The world’s a mess, but I can fake it fine Still breathing—that’s my only line (Hide it), (hide it)... (I'm alive) [Chorus] Romaji: Daijoubu da yo! (Tabun! Tabun!) Dare ni mo barenai dansu wo shiyou Sekai wa hen de, boku mo sou Demo daijoubu sa… tabun ne? Translation:

[Expand Post]

I’m gonna be daijobou! (Maybe! Maybe!) I’ll dance like no one has a clue about me The world is weird, and so am I— But I’ll be okay… probably... right? [Verse 2] Romaji: Boku ni nita kao de afureru kono machi Hen na fuku demo dare mo ki ni shinai shi Nidomi sura sarenai boku ni Tada hitokoto “mata nounai de maigo daro?” Hen na tsubuyaki, waratte nagasu Baka no furi suru hou ga raku sa Mure no naka wa heiwa kamo Baka de utsuru kanji, boku to kimi mo Translation: In this street of faces that look like me No one cares how strange I seem to be No one even stops to take a glance Just one stray voice—“lost in your thoughts again?” I laugh off the weird things people say Pretending I’m dumb keeps stress away In this crowd, it feels less strange Stupid and contagious—just like me. Just like you. [Pre-Chorus] (same as before) [Chorus] (repeat) Romaji: Daijoubu da yo! (Tabun! Tabun!) Dare ni mo barenai dansu wo shiyou Sekai wa hen de, boku mo sou Demo daijoubu sa… tabun ne? [Bridge / Outro Hook] Romaji: “Daijoubu da yo” tte, kikanaide Nantoka katachi ni suru dake sa Translation: Don’t ask if I’m okay—it’s all pretend I’m just holding the shape until the end

>>13245 >>hallucinations Oh sometimes I like when just mumble things for example:

>>13246 >Only my pillow knows how I actually feel >My thoughts are screaming louder than my voice >The world’s a mess, but I can fake it fine >The world is weird, and so am I >Pretending I’m dumb keeps stress away >Don’t ask if I’m okay—it’s all pretend Damn, the lyrics are quite dark. I like it lel. >>13262 >Oh sometimes I like when just mumble things Kek, same. Sometimes the hallucinations amuse me more that a clean result lel. I would wish that these models we more permissive and allowed to change parameters. (Like sampling or guidance). I remember making funny stuff with udio, with prompts like "comedy stand up" or "radio broadcast". Sadly these models are "lobotomized" to only generate generic slop songs. Whatever heres some experimental music.

Nice, I just found your place. I made this song to test out v4.5, hope you guys like it

>>13302 Lyrics: Âme neuve Flamme toute neuve S’enfonce dans le noir Loin du dernier espoir Mais avant que tu plonges si loin je veux te dire Tout au fond des yeux Adieux Et si la brise exauce mes vœux Alors je serai heureux Seule une braise Pauvre petite braise Elle s’est peut‑être tue Sous la voûte sans vertu Et si l’aurore renaît encore Je remercierai les cieux Et quand tu reviendras je serai heureux Âme neuve Flamme toute neuve S’enfonce dans le noir Loin du dernier espoir Mais avant que tu plonges si loin je veux te dire Tout au fond des yeux Adieux Et si la brise exauce mes vœux Alors je serai heureux

>>13288 >Damn is quite dark That's the intention but the original lyrics werent that explicit but I dont know japanese to correct them myself. I liked the second one btw, I usually promt Post rock Progressive Unusual Vocaloid japanese math rock to get interesting stuff but Suno 3.5 and 4 for some fucking reason refuses to give a Hatsune Miku like voice, this was possible with older version. The results sound glitchy as fuck tho. Here's the alternative version of that song i didnt like as much as the other I posted.

>>13310 Need help with Japanese? I speak it almost natively

>>13302 >Nice, I just found your place. Your are welcome! We are a few but we are very active at least i am lol >I made this song to test out v4.5 Oh yeah, i just heard about the model update. How much did it improve? >hope you guys like it Sounds melancholic and mysterious. Interesting. >>13303 >Lyrics: Is it french? cause it looks like it. Also what is it about? >>13310 >>Damn is quite dark >That's the intention Kek, yeah but it got me off guard. I thought it was a happy song due to the intro lol. >I liked the second one btw, I usually promt Post rock Progressive Unusual Vocaloid japanese math rock to get interesting stuff The guy that made those songs uses the tags "Artcore", "Anime ost" and "Boss Battle" a lot kek. Going to post more songs lol. Also Heres the link of the 2º song: https://www.udio.com/songs/wh7mB4MPLFAwH6uwMu1vn7 >Here's the alternative version of that song i didnt like as much as the other I posted. Yeah it sounds more robotic but it still good, I like when his voice changes at the “mata nounai de maigo daro?” part. >>13317 >Need help with Japanese? I speak it almost natively Nice, hey i have a question. Does the songs sound "fake" to you? I have noticed that in some languages (like spanish or russian) the lyrics of the song sound "fake". Its like an uncanny valley for music kek. Sometimes the vocals have the correct accent but the words are just complete nonsense. Does this happens in japanese too?

>>13320 I dont like to samefag nor to spam. But this is somewhat important. i was lurking in the hugging face spaces and found something interesting. A new model (that was created 12 hours ago) called ACE Step. Its a diffusion model of 3.5B parameters (Quite small lel), made by a team of random guys predicted lol >>13245 "someone will eventually make a good model" . I am testing it and its good. It is Quick[/moe](~10-30 secs per gen), has [moe]Quality[/moe](Compared to other open source models) and most importantly Aleast for me kek it has a lot of [moe]Options. (It allows to change things like steps, guidance, seeds, etc...) It also has some extras like "retake" (changes the song), "repaint" (change a specific part), "edit"(change the lyrics or the instruments) and "extend"(change the length of the song). I did some testing, and these are some results. >1. I used the prompt "8bits, synth,120 bpm" and set the steps to 10 and length to 30 secs. It needs more cooking lel >2. I used the same prompt as before but used 60 steps and guidance (loud warning alert) >3. Accidentally made a fnf-like song using the same prompt but with 100 steps and -1 as seed kek >4. I made this one using the prompt "expressive rap song" and some lyrics about oranges. It doesn't sound like rap though lol. >5. Tried using the same prompt as >1 + the tag "boss battle" and it went full "chinese?". The lyrics sound like pure nonsense kek. (Sorry if this songs causes aneurysms geg) Some recomendations: ~100 steps and ~10-20 guidance are the best settings i could found. Also i recommend to try this model before some (((jewtuber))) tell normies about it and the demo collapses lel. Heres the Demo: https://huggingface.co/spaces/ACE-Step/ACE-Step

>>13320 v4.5 audio quality is pretty good, I'm still pushing its limits, no hallucination or shimmering so far

>>13320 Yes the lyrics is in French, it's about innocence lost, broken dream and hope at the end. >>13320 It really depends on the lyrics. It's like English, you can always tell which one is written by bots and which one is written by human, although the line is pretty blurred these days. Which song do you need help with? I can listen to it and give you my opinion

I'm wanting a full version of the little AI song at the end of this video: https://www.youtube.com/watch?v=_HB7rQpfdYk&t=7306s Nice song about the overuse and dependency on technology and AI to do all the thinking for us, no information retention as there's never a need for it. And the "Read my mind and press delete" is a good representation of letting algorithms completely control what you see and do without ever needing to choose for yourself. No need to use the exact lyrics, but here's the transcription if it is useful or wanted: Click, tap, forget, repeat Click, tap, forget, repeat Click, tap, forget, repeat Click, tap, forget, retreat Click, tap, forget, repeat Memory fading, dopamine beat Click, tap, forget, retreat Read my mind and press delete Fading fast, neon dream Too much noise to hear a scream Brain drain, wire chain Only silence knows my name. Click, tap, forget, repeat Memory fading, dopamine beat Click, tap, forget, retreat Read my mind and press delete

Just made this one in v4.5, clean generation, no extension. The song is about Odysseus and the Sirens in case you're wondering.

>>13344 Lyrics: Ulysse, reviens vers nous, laisse‑toi bercer  Odysseus, return to us, let yourself be lulled Par nos voix qui chantent, qui chantent pour ton cœur blessé  By our voices that sing, that sing for your wounded heart Tes batailles sont finies, ta maison au loin s’éloigne Your battles are over, your home drifts far away Nos secrets te seront contés, comme aux vagues le dieu les confie Our secrets will be told to you, as the god entrusts them to the waves Vois nos corps sur les roches, à la lumière de la lune See our bodies on the rocks, in the moonlight N’ai‑je donc pas assez bonne mine ? Do I not look fair enough? Oublie Ithaque et Pénélope, oublie les espoirs perdus Forget Ithaca and Penelope, forget the hopes you have lost Ici la douceur t’attend, pour toujours tu seras sauvé Sweetness awaits you here; forever you will be saved Ne sommes‑nous pas assez belles, Ulysse, dis‑nous la vérité ? Are we not lovely enough, Odysseus, tell us the truth? Nos corps ne sont‑ils pas parfaits, tels que tu l’as imaginé ? Are our bodies not perfect, just as you imagined? Abandonne‑toi à nous, à la mer d’un amour éternel Surrender to us, to the sea of eternal love Ton corps trouvera son repos parfait, dans les profondeurs, sous notre aile Your body will find its perfect rest, in the depths, beneath our wing Tes cordes te retiennent, un regard de haine et de désir Your ropes hold you, with a look of hatred and desire Pourtant, c’est bien à nous que tu veux appartenir

[Expand Post]

Yet it is to us that you long to belong Tes hommes n’entendent pas, ils ne voient pas la beauté Your men do not hear, they do not see the beauty Le supplice de te voir souffrir, à jamais condamné The torment of seeing you suffer, forever condemned Vois ce bateau s’éloigner, nos voix diminuent au vent See that ship sail away, our voices fading in the wind Et la froideur qui monte, car nos pouvoirs s’éteignent lentement  And the chill that rises, for our powers slowly fade L’échec a un goût amer, la beauté s’évapore Failure tastes bitter, beauty evaporates Rejoins‑nous dans les profondeurs, avant que nous ne soyons plus qu’un souvenir  Join us in the depths, before we are nothing but a memory

>>13332 >v4.5 audio quality is pretty good, I'm still pushing its limits, no hallucination or shimmering Good. It looks like the lyrics have improved and sound more natural. >>13333 >Yes the lyrics is in French, it's about innocence lost, broken dream and hope at the end. Wow, your songs are very Intriguing. Meanwhile the better i can do are instrumetals kek. >>13333 >It really depends on the lyrics. It's like English, you can always tell which one is written by bots and which one is written by human Fair enough, I have noticed that lyrics and music quality are detrimental to eachother. A song with good quality has usually bad lyrics, and a song with good lyrics has bad sound effects. Still looks like suno is in a good balance. >>13333 >Which song do you need help with? I can listen to it and give you my opinion Oh i dont need help, but thanks. I usually dont generate songs wth lyrics. >>13334 >I'm wanting a full version of the little AI song at the end of this video Maybe try extending it, uploading it to suno or udio and use the "remix/cover". Although i think you need a pro plan to do that. Still i am gonna try to make that song using ACE. bad news, the space in hugging face suffered a runtime error, so no gens kek >>13344 >Just made this one in v4.5, clean generation, no extension. >5 min song. Wow, thats impressive. How long are the gen now? How much you had to wait for it? Also theres some good news. The model card has been updated and theres some info now. ACE step is actually a mix of three models (a diffusion-based generation + Sana's Deep Compression AutoEncoder (DCAE) + a lightweight linear transformer). But the most important part is that it is ligthing fast. "15× faster than LLM-based baselines (20s for 4-minute music on A100)" and it can run on graphic cards like Nvdia. Here a table: [code]"Hardware Performance: Device 27 Steps 60 Steps NVIDIA A100 27.27x 12.27x RTX 4090 34.48x 15.63x RTX 3090 12.76x 6.48x M2 Max 2.27x 1.03x"[/code] Also going to post a gen i did yesterday and another song of Lunaris. Now I dont have more music. send help kek

>>13333 Check

>>13317 It's nice to have someone to check out the lyrics that Chat GPT redone, does this sound good or natural to you? >>13188 I'm talking about the lyrics not the voice I originally writte it in english but didn't like the results Suno give, (maybe because I'm a fucking weeb lol). >>13320 Damn the second sounds quite good I should try Udio again because when I try it a year ago wasn't that good. >I have noticed that in some languages (like spanish or russian) the lyrics of the song sound "fake". I HATE AI music with Spanish Lyrics maybe its because all the examples I heared before are lame slop prompted by people who writte the most cringy lyrics ever but I can't listening to it man. Some Instrumental track I made today, I really really wanted to repeat the hallucination chorus from the beginning after 2:15 but no matter how much I tried the AI keeps doing something different or just ending it with an outro.

>>13421 >I HATE AI music with Spanish Lyrics Spic here, I do hate music in spanish overall.

>>13361 >Wow, thats impressive. How long are the gen now? How much you had to wait for it? No idea why but v4.5 is super fast on my machine, I'm talking like less than 20 seconds for a song. Even older models like v4 or v3.5 can take it up to 1 minute or so

>>13421 If possible, you should look into the syllables structure (mora) of Japanese lyrics. Most of the time the sentences are not even, and in modern songs, the sentences are often shorter than their 20th century counterparts too

>>13421 The lyrics >>13188 >>13246 definitely sounds AI generated. You should break down your sentences more. For Japanese song you should write the lyrics entirely in Japanese, NOT romaji, and while prompting, you should convert the lyrics into hiragana or katakana only for phonetics. I've seen AI messed up kanji pronunciation a lot of times, it might sound normal to non-speakers, but it's a dead giveaway to native speakers.

夏国の彼方 (Natsukuni no Kanata) Beyond the Summerland [Verse] 蝉声止む畦道 | The levee path where cicada song falls silent 水色空に声返す | We call our voices back into the water-blue sky 君と西瓜分けた夜 | That night we shared a watermelon, you and I 蛍光る田んぼ眩し | Fireflies flashed; the flooded fields were blinding [Verse] 君探して影の中 | I search for you amid the drifting shadows 稲風渡り名を呼ぶよ | Rice-field winds carry your name across the rows 答えなく蝉に響く | No reply, only cicadas answer back 日暮し夜へ導くと | Dusk itself leads onward into night [Pre-Chorus] 風はささ | The wind rustles 遠く光 | A far-off light 永遠の夏が呼ぶよ | Eternal summer beckons her away 手を伸ばす | We reach out 影踊る | Shadows dance 守れぬ光へ行くよ | She walks toward a light we cannot guard [Chorus] 遠き夏籠る闇へ | Into the dark where distant summer hides 我らの歌響き乾く | Our song rings out and dries upon the air さよならは言わず探す | We speak no farewell, we just keep searching 蒼い水まで追いかける | Chasing her down to the deepest blue water [Alt-Chorus] 夏国から蒼い音 | A blue note drifts from the land of endless summer 君は掌 光の中 | You, a small flame cupped within that glow 我ら届かず風切る | We cannot reach you, the wind slices between 眠る森のほとりに歌 | And our song lingers by the edge of the sleeping forest

>>13431 >>13433 Lyrics break down as prompt: [Verse] せみごえやむあー ぜみち [Instrumental Bridge] みずいろそらにー こえかえす [Instrumental Bridge] きみとすいかわー けたよる [Instrumental Bridge] ほたるひかるたー んぼまぶし [Instrumental Bridge] [Verse] きみさがしてかー げのなか [Instrumental Bridge] いなかぜわたりー なをよぶよ [Instrumental Bridge] こたえなくせみー にひびく [Instrumental Bridge] ひぐらしよるへー みちびくと [Pre-Chorus] かぜわー ささ とおくー ひか えいえー んのな つがよー ぶよ

[Expand Post]

[Pre-Chorus] てをのー ばす かげおー どる まもれー ぬひか りへゆー くよ [Chorus] とおき . なつこ . もるや . み へ われらのうたひ . びきかー . わ く さよな . らわい . わずさ . が す あおいみずまで . おいかー . け る [Chorus] なつく . にから . のあお . い ね きみわてのひら . ひかりー . な か われら . とどか . ずかぜ . き る ねむるもりのほ . とりにー . う た

>>13421 >Damn the second sounds quite good I should try Udio again Yeah, looks that the model has improved, the lyrics sound much better now. >I HATE AI music with Spanish Lyrics maybe its because all the examples I heared before are lame slop Kek, yeah the problem is that a lot of people dont know how to prompt and also have bad music taste lel. >>13433 Very comfy and genuine song. I like the slow pace and the melancolic vibe. >>13361 I tried to generate more songs but the ACE space has imploded due to a huge amount of normies using the servers lol. Here some songs i found in udio though, the background music is bad but the lyrics are almost perfect (aleast for my ears lel) Here the guy who made (1) and (3): https://www.udio.com/creators/Melanintoninman. I also have noticed that the weight of the songs has doubled. Still worth it for the huge increase in quality Also someone knows what the second songs says? Heres the link: https://www.udio.com/songs/qvAn2CBwndhckUXAyUYABH

>>13444 Thank you, I'm glad you enjoyed it.

>>13444 more than the prompts the lyrics are VERY cringy Anyways heres some stuff I did today maybe the intro is familiar to some of you lol