Bad Apple but AI Generated (used MMD as reference for it) 

I just want to mention that this was made purely to see if stable Diffusion could make it look more like anime instead of CGI. In no way is this an attempt to make it "better". I love both the original black and white, and the 3d animated version (which this used as a base when generating) and in no way am I trying to be better or replace them. I thought this was a fun and interesting thing to try. I kinda feel like the attention and views this video got should’ve gone to the original MMD (top of description) so please go check that out and like it if you see this. Personally I feel like this video is a little too close to the original MMD and would’ve preferred if it was more unique. Also, I am making no money whatsoever on this video. Ads are because the song is copyrighted, so all revenue goes to the copyright holders. I completely understand the issues people have with AI. I personally don't feel I am doing anything unethical with this video, but you're free to have your own opinions on that. I expected this to get like 20 views like my other stuff. There was no ill intent in the creation of this video. Thank you for watching and I hope you have a great day! Also, regarding a tutorial on how I made this, it's been so long that honestly I've forgotten half of how I made it. And as of now there are so many new tools to use so the tutorial would be very outdated. Maybe in the near future I can make videos showing the process of experimenting with stuff like this and show the full process of how I make it, using new tools and strategies. Might soon try an AI version from the black and white version. I think the technology exists now to do a decent (probably still janky) version. But it will be more of a full video with me experimenting and explaining what I’m doing. Just need to make sure I’m doing my college work first. - edit from 9/8/2023 The rest of this comment is from a while ago. Thank you so much for getting this to nearly 100k views! That’s insane! Bro what now 1 month later it’s at nearly 200k Thank you so much! I get like 25 views on my other stuff lmao

as a non-tohou player, this is the first time i have any idea what any of these characters are intended to look like outside silhouette form.

this animation will never die

Love how the AI thought Yuuka's umbrella at 3:03 was the moon.

0:53 The ai made her wink and smile was adorable.

This feels like a "Bad Apple but I dreamed it after spending too much time looking at Touhou fan art"

0:56 Og: scary Ai: kawaii

The fact that the AI generation (which I believe makes the animation frame by frame without trying to create a link with the previous frame) clicks the character’s design during the movements is surprisingly very smooth, I love it!

2:46 is a moon jumpscare.

The remaster we deserved. But seriously, this AI coloured animation is finally looks somewhat closer to original fluid movements and transitions!

0:27 Bruh Became Human 💀

This is gonna take awhile but here we go 0:01 Reimu Hakurei (I like how sometimes she Change expression’s and of course not too sexy side view ,since the 3D/MMD ver has her side boob exposed luckily the Ai actually censored it in a way) 0:15 Marisa Kirisame ( she has a cute happy expression , but i think she’s missing her leg / invisible leg) 0:29 Patchouli Knowledge ( this is not bad but of course the ai still messed up the hands In some frames) 0:35 Remilia Scarlet ( Lego hands lol 😂 but In a way her hands still in place just the ai having trouble with the object is hilarious) 0:43 Sakuya Izayoi ( I don’t find any problems on her so far , as expected from the “perfect” maid) 0:50 Flandre Scarlet ( she’s actually gives you a cute look instead of the whole psychopath look which make sense now since she’s no longer the psychopath people think she is , just a bit destructive and lonely ) 0:58 Youmu Konpaku (her clothes keeps changing colors and she did 360neck turn..dam that’s creepy) 1:04 Yuyuko Saigyouji (she actually almost hard to see on the youmu sword scene until the camera focus on her and in some frames she changes her eye color , overall not to messy ) 1:11 Komachi Onozuka(she only gets messed up if the camera isn’t too focused on her , overall nice shading by the AI) 1:18 Shiki Eiki ( just like komachi she gets the special shading ) 1:24 fujiwara no mokou(her hands in some frame is funny and also her head at the end looks weird but overall it’s insanely good) 1:32 keine and her hakutaku form ( they exchanged glasses lol) 1:38 Eirin (i think she stared at my soul) 1:46 Kaguya Houraisan ( her objects got screwed up a bit but overall it’s good) 1:53 primsriver sisters ( in some frames they change colors but the object doesn’t get messed up that much ,good) 2:00 - 2:06 Chen , Ran , Tewi , Reisen , Momiji (all of them looking good ) 2:08 Sanae Kochiya ( she’s actually got the special shading which is cool I guess) 2:11 Hina ( not much of a change really) 2:13- 2:19 Kanako and Suwako ( holy this is good I like how kanako doesn’t express that much and suwako happy face is cute) 2:20 - 2:35 yukari yakumo , tenshi hinanawi(they actually looks so good woah) 2:37- 3:34 aya , suika , alice , nitori , yuuka , Elly , pc-98 reimu and pc98 Marisa , windows marisa , windows reimu( they all look so good in their own way , maybe suika having mouth problems and pc-98 reimu and pc-98 marisa turning into windows in some frames) Man that took like 20 minutes 😂

Epic improvement

the fact the ai picked on the the designs so well is crazy also 1:01 the ai did have a problem and switch between making her head look like a shrine cause of her moving around so far back

Imagine if a competent artist did this but more so used it as an animation tool and touched up every frame to make it make sense and flow better, it would be so awesome.

3:03 like how her umbrella turns into the moon somehow

再現度高すぎて、なんか懐かしくなって泣けてきた 子供の頃からずっと見続けてる

the speed at which ai technology is progressing is scary

I love that Flan actually winked.

AI generated videos are like what it's like to dream or try to remember a dream but it's not vivid enough of a memory

Not being in the touhou fandom it's cool to know what these characters look like.

The way the colors of the character's eyes shift slightly is mesmerizing to look at. Not 100% in the touhou rabbit hole but if I'm not mistaken for some of those shifts was cool cause the character is sometimes drawn with either of the 2 eye color options. If am wrong on that is still very well done as most of their designs are pretty on point throughout the frames despite it being ai generated.

I love how the AI was able to recreate the subpar animation on the 3D rigs EDIT: I wrote this to be harsh on purpose for the comedic effect. Considering the circumstances the original animators were working with I can understand the quality. It's great for what it is :) Take a chill pill yall!

3:32 I absolutely love that the ai made the dark side the moon and the light side for the apple, it just adds more to the themes of the animation. Also I love the fact that every character had facial expressions that matched their personalities (the original literally had no eyes to base anything off of). The transitions were surprisingly nice as well, it seems ai has a weird way of effectively doing it better than most people when using color.

0:56 That wink was adorable!

pause right at the beginning at 0:01, (I'ts a weird glitch!)

2:28ここの天子 めちゃくちゃ天子らしくて凄い

okay this is just my fever dreams with how it constantly shifts and evolves lol

I'm sure the original creator would appreciate if you linked the MMD version in the description!!

1:55 blowing the violin and bowing the trumpet how nice

this Flan's wink got me fr

影絵→CG→AIイラストという一連の変化に時代の流れと技術の進歩を感じずにはいられないよね

I like the fact that the animation style is always changing, its more abstract, like the original black/white video

the ai wobbling shapes and colours around with every frame reminds how these same characters have been depicted in thousands of different artstyles and fanons. how poetic.

THIS ANIMATION BRINGS ME LITERAL CHILLS

0:22 from an evil mansion to a Half life 2 beta looking combine citadel, epic

I kinda love this style actually. The way the textures and shading details change randomly but smoothly. Great visual for a music video!

This is the best animation I've ever seen for AI Generated

0:46 MY FAVORITE

The fact that it’s like weird 2.5D models is so cool!

2:11 i got goosebumps

Bro. I had no idea what any of these characters looked like before this. But the only thing I got right was the chick who shot the finger gun had purple hair.

Weird how this is one of the most unsettling versions i've ever seen. Adding color and definition detracts from so many of the stylish transitions; genuinely interesting to see how much the monochrome silhouettes truly make Bad Apple as great as it is

0:22 That building was so unexpected I nearly choked lmao

1:02 scares me

This is the first time I actually see the characters! Thank you! 🤩

now we just need an animator to reanimate the entire thing and it'll be *french kiss* 🤌

色がキラキラかわるのかっこいい…

1:02 ayo, her head broke!

I like how some eye colors change, the two I could actually eye though were Yuyuko and Tenshi

Wow never seen AI good actually some AI just tend to make curse in the video but in this one it was smooth and perfect and kind of smooth like dang.

Y'know, ironically, this gives the video more life than it ever had. I never really viewed the characters as overly distinct.

1:41 AYO 💀

Yuyuko's eyes: 🔴👄🔴🔵👄🔵🔴👄🔴🔵👄🔵🔴👄🔴🔵👄🔵🔴👄🔴 All jokes aside this is the best Bad Apple in color so far that's pretty consistent AND really nice to look at. I really like it! Flan so cute

guys, this version uses the 3D colored version of Bad Apple as a base, that’s why it is colored an coherent

This deserves more view. I can't wait to see a fully colored version and polished version of the original animation!

The improvements are very visible this time, and I can get why I see so many illustrations from danbooru that have been uploaded throughout the years. Becuz the AI program was trained on Pixiv and Danbooru. Though a nice effort for compiling the video :)

3:38 language of the gods

This would be a coloured remaster if it was actually animated and did correctly.

すごく自然！！！めちゃくちゃ好き

0:55 this part is really cuty idk why 😊

absolutely perfect!!! wooow, new record!

i think the classic bad apple is perfect as is, it cant be "improved", that said, this version is quite fun too, one interesting thing about the original is that its a bit abstract since you cant know how the characters would look like in color, on this version, on the other hand, you know but at the same time you dont, its abstract despite the fact that you can see, since the shape/colors change/morph all the time and your brain cant really process what he is seeing.

Flandre's part is an ABSOLUTE. MASTERPIECE.

Flandre's smile is cute rather than creepy in this version. 0:55

I’m glad stable diffusion is free to use and open source giving us and developers the opportunity to super elevate AI

The animation is so smooth!

天子と紫のパートかっこよすぎる 鳥肌立った

God damn this is mind blowing

I like how the characters looks keeps changing, makes it feel like a dream

This is actually nice.

Wait a minute! This film is actually stitched together frame by frame? If it's so smooth, how long did it take you to make it?😲😘

The Guy who made this video: I have achieved SUPER CONSISTENCY. I have achieved the power of God and Anime

Mad respect for the amount of expertise and work it must have took to get it to look so good. I'd love to see more about how it was made.

People love to complain about real human beings stealing jobs from them even though they need them to live, but when people can make billions of robots to do every creative job that humans would actually want to pursue in life people will defend it existing and I don’t understand why. When every creative job comes down to just clicking a few buttons what will we pursue now? Shouldn’t we be making robots do the technical work nobody wants to do?

🙃 Loving how people are hyping this up! So I'll say it. Ai art is theft. The person who created this video did no work. They're profiting on other artists work.

0:59　幽々子さん　てんどんマンになったりロボットになったり城になっててくさ

Lol, this is fun, but you can definitely see the jank here from the AI. 1:02 made me laugh when she just casually snapped her neck.

WhaT. THIS IS A MASTERPIECE, really nice

POV: you said the edibles ain't shit

That's flandre was the most adorable thing I've seen

So good, you are genius. Keep going

That's gonna be a really cool timestamp of AI capabilities come ten years.. K n o w i n g i n t e n y e a r s , B a d A p p l e w i l l s t i l l h a u n t u s w i t h r e c r e a t i o n s . . Nah but actually, this is a cool lil project. :>

There's something disquieting about the constant influx abd changes in the images that somehow makes the whole thing work for me. Really nice job.

Bad apple, now in full technicolor. It can can use some streamlining, but damn this good

0:55 Why does it look like Flan is breaking the 4th wall and singing along?

Ive never seen touhou except for a few things and the music, so it’s cool to see what the girls in the animation actually look like.

The first actually good piece of content from ai

Fui de los primeros en escuchar esa buenísima canción y ver lo que la IA hace xD saludos mano ✨

Damn, this may be the only bad bad apple version

The animation is smoother than my life

I love how round objects just transform into the moon! XD

this has no right being this good

bad apple but they turned the lights on

fat upgrade than some of the bad apples

Now that we have AI video generating features, you should try that now to see how accurate it does. Lol Overall, this is still impressive

never underestimate AI just because it cant make real hands

Me eating an apple watching this: My brain: THROW IT ON THE GROUND

I like how Flandre winked at us!

Flandre smiling and winking instead of looking menacing is the most adorable thing.