AI made it with a lot of human feedback :D The songs are made in 32 second intervals and I choose the best sounding ones. It's like catching lightning in a bottle. I wish there was some secret to it, but it's just a game of chance. I'd say 75% of those 32 second intervals go to the trash, 20% are passable and 5% are golden. If you have enough patience, you can make really nice songs yourself.
Vibes might be nostalgic, but this was made only a few months ago and would have been impossible to make a year ago :D I hope AI tools will make it possible to turn nostalgia into reality.
Oh most definitely, my favorite part is when the AI extrapolated 0010101110001010011001110101000 into "kiyottsukete ne umi wa shiranai" instant classic. This is what digimon with poor work/life balance unwind to with with a glass of data whiskey and a digital cigar, after a long shift saving the digital world
I had a feeling the album art looked somewhat modern, but I was convinced the music was actually from the 70s! This is the first AI-generated song I’ve ever listened to and its amazing! I was going to look for the rest of the album lol
I'm glad you like it, but I'm also sorry because now your standards have been set too high. The quality of AI songs is very inconsistent. There's no rest of the album, just one more song. There'll be two more coming in December.
@@YellowAiOrchestra nice, i’ll have to check those out as well! I probably won’t “intentionally” look for more AI-generated music, but i’m still happy to have found this song nonetheless. Of course, theres already plenty of great original music out there from this era to listen to, but I think its still worth giving AI songs (with this quality) a listen every once and a while.
That's a great question. I usually think of the most important parts of the lyrics, then I use ChatGPT, Gemini or one of the numerous local models to expand upon them and adapt them to Japanese language. I'd make songs in more languages, but Japanese works best for some reason. Even English sounds like gibberish 70% of the time. Heck, Japanese does too. Japanese is a great language for singing in general. You have a lot more control over vocaloids, except their voices are limited. Vocaloids have set, recognizable voices. AI voices always sound different so they don't plagiarize real artists' voices. Using audio models, you paste the lyrics into the prompt and hope for the best. The entire song doesn't just pop out on its own. You have to carefully curate every verse in 32 second intervals. Compared to the vocaloids, there's less beatboxing, but there's way less control. I was never into vocaloids. There's only one vocaloid song I've ever liked. In summary, I have very little control over what goes on in the songs. What I upload is the best Udio 1.0 and 1.5 have to offer when it comes to city pop. I'll upload some metal and punk songs soon, so you will hear how the AI imitates those genres. I am not certain about the future of AI music since there are literally only two services that provide those: Suno (better music) and Udio (better voices). Since they're services, and not local, they can be taken down or go behind a serious paywall, this little tech demo of mine could face an untimely end. But worry not! I have a perfect farewell album for when that happens! :D
BTW AI doesn't just invent something from nothing, in a sense there is no real AI, but they are just advanced bots that are able to interpret existing songs and such and generate new songs from them.
@@ziongite I know. The models are fed with thousands of real songs. I'm not sure if this is the same thing as bots since even the silliest bots I chat with to are more enjoyable to work with than Udio prompt input which spews out nonsense 9 times out of 10. :D
Not sure if you deleted my comment or I mistakenly did, but I was just saying how great it is to see some other Japanese AI music. I created a City Pop album myself earlier this year that I just uploaded to my channel if anyone wants to check it out.
Did AI come up with the strange song title「海に飲まないで」? It's grammatically odd. In English, it's "Don't drink to sea (or ocean)". Shouldn't it be "Umi ni Nomare naide"? (Don't be swallowed by the sea). Waza to 「海に飲まないで」ni shitandesuka? PS : Great song, by the way 😺
Oh, God! Well, it's too late for changes now. I am sorry. Japanese is my third language and most things I generate are translated threefold. The AI came up with the title. In my native language, the sea swallows you (or devours you it's the same, really), so I thought it'll be the same in Japanese. In my mind, the title should be "Don't allow the sea to devour you." You know how young men are naturally attracted to dangerous places, especially the rails... There's also an element of sea sirens/mermaids luring foolhardy boys to their demise... Ichika should be singing how there's no going back, the nature doesn't care about the boy in question. My mother tongue is very specific, while Japanese (especially when it comes to song lyrics) feels metaphoric. I've checked at least twice that the kanji I need is the same for drinking, and to me, the sea swallows people like it would swallow a bite of food, it's not taking a sip of people... Well, you can't drown unless you drink too much water... Idk, I'm a total noob. When I make songs, I write in the heat of the moment using free verse and go along with them ASAP. I think in one language, feed the prompt in English to adapt the song in Showa-era Japanese and to translate back so I see what it says. I'll make more songs in English in the future. I'll make them in my language too if the AI models start supporting it. This song is based on a real life tragedy in which a friend of mine lost her brother on a cruise ship, his body was never recovered.
@@YellowAiOrchestra Naruhodo. I'm still impressed that even with your limited Japanese language skill, you created such a convincing Showa-era song. It's so uncanny that i love it! ❤ The strange song title adds to the uncanniness. It still takes a lot of your blood, sweat & tears to make these songs, so ganbatte ne! 😺
@@brianflynn5355 Thank you! I used to teach myself Japanese by translating Showa-era songs in college. Back then I had to do these things manually. :-)
いい雰囲気
古い雑誌の記事のような感じ。ネットで検索しても引っかからないような歌手。それが切ない
「海「に」飲まないで」の「に」の引っ掛かりが良い
It really does sound like the showa era
It makes me forget music from other eras exists :D
This is a Beautiful artificial Masterpiece.
It's amazing what the technology can produce. There's more to come.
SO GOOOOOOD
Absolutely wonderful 💎
How did AI nearly make a city pop classic????? This is unreasonably catchy! Especially 1:43 like omg dude. I still cant believe AI even made this
AI made it with a lot of human feedback :D The songs are made in 32 second intervals and I choose the best sounding ones. It's like catching lightning in a bottle. I wish there was some secret to it, but it's just a game of chance. I'd say 75% of those 32 second intervals go to the trash, 20% are passable and 5% are golden. If you have enough patience, you can make really nice songs yourself.
It was nice to find such an amazing song.
Wow. I love it. Goosebumps. Legit use of AI bring those nostalgic vibes back to life.
Vibes might be nostalgic, but this was made only a few months ago and would have been impossible to make a year ago :D
I hope AI tools will make it possible to turn nostalgia into reality.
Amazing song !❤
Cara isso é muito bom!
one love. just one love.
しゅごい😂
アルゴリズムのジャケ写に誘われて来てしまいました
なんと「AI」なんですね😱すごい時代ですね～
ちょっと佐井好子っぽいかなって思っちゃいました
調べても詳細が出てこないのでもしやと思ったらAIなんですね……普通に良い曲でびっくりしました。AI凄い!
신기하네 진짜 70년대 노래같아요
Love it 😎
Que canção lindíssima!!!
This song FUCKS, great job dude, cant wait to see what you cook up next
There'll be 15 more songs by the end of the year. I've uploaded some 100 songs by now, you can listen to them if you want. Have fun!
Oh most definitely, my favorite part is when the AI extrapolated 0010101110001010011001110101000 into "kiyottsukete ne umi wa shiranai" instant classic. This is what digimon with poor work/life balance unwind to with with a glass of data whiskey and a digital cigar, after a long shift saving the digital world
@@khaotik1I'd probably never even hear Japanese music if it weren't for Digimon during my formative years. They've had awesome soundtracks!
They really did!!! God, the childhood memories
I had a feeling the album art looked somewhat modern, but I was convinced the music was actually from the 70s! This is the first AI-generated song I’ve ever listened to and its amazing! I was going to look for the rest of the album lol
I'm glad you like it, but I'm also sorry because now your standards have been set too high. The quality of AI songs is very inconsistent.
There's no rest of the album, just one more song. There'll be two more coming in December.
@@YellowAiOrchestra nice, i’ll have to check those out as well! I probably won’t “intentionally” look for more AI-generated music, but i’m still happy to have found this song nonetheless. Of course, theres already plenty of great original music out there from this era to listen to, but I think its still worth giving AI songs (with this quality) a listen every once and a while.
Good quality to sample this
Good luck and have fun!
@YellowAiOrchestra than
オッサンの自分にも刺さると思ったら、80年代回帰の作品なんだ。
バックがロックではなくてフュージョンスタイルなのが当時の流行りだったもんね。
今思えばＪ－Ｐｏｐが都会的でクールになった時代でもあった。
ご視聴ありがとうございます！
自分も長年80年代のJ-Popを聴いてきましたが、特定の音楽スタイルが支配的だったわけではないと思います。ただ、自分の耳には「ビッグバンド」の音が初期80年代のポピュラーなサウンドの基盤だったように感じます。
80年代の日本のポップスを思い浮かべると、シンガーが大規模なオーケストラに支えられている姿を想像します。専門家ではありませんが、当時はテレビ局やラジオ局に自前のオーケストラがあって、それを活用するほうが、乱暴なロックンロールのセッションミュージシャンを使うよりもコスト的に安かったのではないかと想像しています。そのため、楽曲もオーケストラに適したものが多かったのかなと思っています。
AI no umi ni nomikomareta anniv. (2024.11.14)
am impressed but also disappointed at the same time that this is Ai GEn
Eh, it is what it is :D
@@YellowAiOrchestra it is what it is yea¯\_(ツ)_/¯
ルパン三世の挿入歌にありそうすね
かもね
The images are also created with artificial intelligence, I like the image in this video.
Thanks! It's a cool picture. I think it took around 20 gens until I got it right and then I've done some inpainting.
you mention in your bio that you’re a virtual band! are these songs completely ai generated, or is a little more akin to something like vocaloid? :)
That's a great question. I usually think of the most important parts of the lyrics, then I use ChatGPT, Gemini or one of the numerous local models to expand upon them and adapt them to Japanese language. I'd make songs in more languages, but Japanese works best for some reason. Even English sounds like gibberish 70% of the time. Heck, Japanese does too.
Japanese is a great language for singing in general.
You have a lot more control over vocaloids, except their voices are limited. Vocaloids have set, recognizable voices. AI voices always sound different so they don't plagiarize real artists' voices.
Using audio models, you paste the lyrics into the prompt and hope for the best. The entire song doesn't just pop out on its own. You have to carefully curate every verse in 32 second intervals. Compared to the vocaloids, there's less beatboxing, but there's way less control. I was never into vocaloids. There's only one vocaloid song I've ever liked.
In summary, I have very little control over what goes on in the songs. What I upload is the best Udio 1.0 and 1.5 have to offer when it comes to city pop.
I'll upload some metal and punk songs soon, so you will hear how the AI imitates those genres.
I am not certain about the future of AI music since there are literally only two services that provide those: Suno (better music) and Udio (better voices). Since they're services, and not local, they can be taken down or go behind a serious paywall, this little tech demo of mine could face an untimely end.
But worry not! I have a perfect farewell album for when that happens! :D
@@YellowAiOrchestra ohh, I see. thank you for answering my question!
@@AdrienDC Anytime!
BTW AI doesn't just invent something from nothing, in a sense there is no real AI, but they are just advanced bots that are able to interpret existing songs and such and generate new songs from them.
@@ziongite I know. The models are fed with thousands of real songs. I'm not sure if this is the same thing as bots since even the silliest bots I chat with to are more enjoyable to work with than Udio prompt input which spews out nonsense 9 times out of 10. :D
Not sure if you deleted my comment or I mistakenly did, but I was just saying how great it is to see some other Japanese AI music. I created a City Pop album myself earlier this year that I just uploaded to my channel if anyone wants to check it out.
I never felt the need to delete any comments, mate. Good luck with the music!
wait so this is AI ??????
It is!
Sleepy
どうせヴェイパーウェイブを揶揄した最近のシンガーだろ😁そうおもって聴き始めたら本格的な声量のある実力派歌謡曲ボーカルアーティストと良曲でノックアウトされ、どんな人だと検索かけたら誰でもなくて驚愕しました…ヤラれたわー✌️
ごめんなさい、騙されちゃいましたか！このチャンネルのすべてが人工的なものだって、結構わかりやすいと思ってたんですけどね :D 来週、演歌の曲をいくつか公開する予定なので、楽しみにしていてくださいね！
일본은 이미 '소녀문화'가 앞서 있군요..
Konnichiwa.🌞
Ongaku ,yasashi no kawaidesu.
Trujillo Perú ,karakimashita
How?
What would you like to know?
@@YellowAiOrchestra I'm just surprised that artificial intelligence can do such a thing.
@@Anteha I don't know either, but I do know it takes a lot of cherry picking from the user side.
昭和と思ったら令和の曲なのか
令和の曲です。
not sure why people in the comments think this is good. stop generating AI slop and go outside
I guess people listen to what they like or move on instead of seething. Cheers!
i know you're trying to be smug about this but it changes nothing
Humanity hates itself so much it won't be satisfied until it is completely replaced by computer program bots
Did AI come up with the strange song title「海に飲まないで」? It's grammatically odd. In English, it's "Don't drink to sea (or ocean)". Shouldn't it be "Umi ni Nomare naide"? (Don't be swallowed by the sea). Waza to 「海に飲まないで」ni shitandesuka? PS : Great song, by the way 😺
Oh, God! Well, it's too late for changes now. I am sorry. Japanese is my third language and most things I generate are translated threefold. The AI came up with the title.
In my native language, the sea swallows you (or devours you it's the same, really), so I thought it'll be the same in Japanese. In my mind, the title should be "Don't allow the sea to devour you." You know how young men are naturally attracted to dangerous places, especially the rails... There's also an element of sea sirens/mermaids luring foolhardy boys to their demise... Ichika should be singing how there's no going back, the nature doesn't care about the boy in question. My mother tongue is very specific, while Japanese (especially when it comes to song lyrics) feels metaphoric. I've checked at least twice that the kanji I need is the same for drinking, and to me, the sea swallows people like it would swallow a bite of food, it's not taking a sip of people... Well, you can't drown unless you drink too much water... Idk, I'm a total noob.
When I make songs, I write in the heat of the moment using free verse and go along with them ASAP. I think in one language, feed the prompt in English to adapt the song in Showa-era Japanese and to translate back so I see what it says. I'll make more songs in English in the future. I'll make them in my language too if the AI models start supporting it.
This song is based on a real life tragedy in which a friend of mine lost her brother on a cruise ship, his body was never recovered.
@@YellowAiOrchestra Naruhodo. I'm still impressed that even with your limited Japanese language skill, you created such a convincing Showa-era song. It's so uncanny that i love it! ❤ The strange song title adds to the uncanniness. It still takes a lot of your blood, sweat & tears to make these songs, so ganbatte ne! 😺
@@brianflynn5355 Thank you! I used to teach myself Japanese by translating Showa-era songs in college. Back then I had to do these things manually. :-)