3D Anime Analyses and Breakdowns

Hello,

Today, I wanted to post observations that I have made while watching various 3D anime, and I hoping that others could contribute to it as well. Please move this to the correct forum if this is not in the right place. Also, if this has been discussed before, please post a link on here, as there was too many topics to look through when I searched “anime.”

##I guess people don’t like breakdowns to be too long and leaving out details, so I added a TLDR to this##

DISCLAIMER: These are all analyses that I have made myself, so I could be wrong, or cannot explain it as clearly as someone else can. Since I am a character modeler, most points that I will be talking about have to do with character design and animation. SPOILERS AHEAD.

Table of Contents

  • Intro
    [LIST]

  • Listing which anime that I’ve studied the most as of this posting.

  • Analyses and Breakdowns of each animation

  • Since there is a character limit, I have to move the overall animation breakdowns here.

  • Character animation

  • Looking like a robot is a major problem that most of these animation studios have. While in 2D characters can stand still and it won’t break the scene, it doesn’t translate well to 3D. Other times, there’s just a little too much movement going on with a character. We’ll be looking at walking/running animations, body language, and gestures.

  • Shadows

  • Some studios have different approaches to their shadows. Some look better, some look worse, some are just personal preference. I’ll discuss cast shadows, shadow maps, and shadows built into the texture.

  • Facial animation

  • The facial animation depends on the studio. Some will go realistic while others will blatantly try to do 2D faces. Nothing wrong with either as long as they are done correctly. However, lip-syncing is where most animations kill their characters. Lip-syncing, facial expressions, and textures/shadow maps used for emphasis will be discussed.

  • More to be added

  • Please make suggestions, or try to add to any topic.

[/LIST]

So, for this topic (and possibly any future topic that I make on this subject) I will use these animation studios and their works, purely for the fact that I have seen only these as of this posting.

Movie - Bold
TV Series - Normal
Commercial - Underline
Other - Italicized
##Newly Added## - ##Number Sign##

Toei Animation
-Expelled From Paradise https://kissanime.to/Anime/Rakuen-Tsuihou-Expelled-From-Paradise

Polygon Pictures (As of right now, both animations are available on Netflix)
-Sidonia no Kishi (Knights of Sidonia) https://kissanime.to/Anime/Sidonia-no-Kishi
-Ajin https://kissanime.to/Anime/Ajin

Sanzigen
-Aoki Hagane no Arpeggio: Ars Nova Cadenza (Arpeggio of Blue Steel) https://kissanime.to/Anime/Aoki-Hagane-no-Arpeggio-Ars-Nova-Cadenza/Movie-Full?id=124411

Kyoto Gakuen University
-A Butterfly’s Dream https://www.youtube.com/watch?v=IB9C3FHotLA
- Breakdown https://www.youtube.com/watch?v=qQUFe3y1JXY

I hope that you enjoyed this first post. I’ll try and follow the table of contents that I created above, so expect that next time to come up.

Expelled from Paradise

TLDR: Look at Dingo for textures, look at Angela for animation

There are many noticeable changes in how the characters appear on screen in Expelled from Paradise. If you want to see many examples in just one character, then pay close attention to Dingo. The other male characters in the movie are good examples as well, but they don’t get as much screen time as he does. You’ll notice that they put more work into Dingo’s texturing and shading when compared to Angela’s, but at the same time they put more effort into Angela’s animation when compared to Dingo’s. His face and chest rely more on the shadows that were drawn into his textures. You’ll notice this a lot if you pay very close attention to his face and see that most of the time the shadows on his nose are a texture. As for his animation, it never seems to change that much. He blinks, he can have wide grins, and he actually moves his jaw, especially when seeing his face from the side or behind. However, that’s pretty much it. This is understandable because of his personality; I’m just saying that you won’t be able to analyze him for facial animation as much.
Angela is quite different than Dingo. She relies more on animation than Dingo does. She is the main heroine, an adult woman in a 16-year-old body, and (let’s face it) this is Japan; She HAS to move around more. The advantage to this is people can study her more for physics (though some places more than others) and facial animations. Angela shows a wide range of emotions from anger, rage, happiness, sadness, and anything else in between when compared to Dingo. She is also more expressive, so you can watch her body for animation references as well. Her shading is similar to Dingo’s, except she’s not as reliant on textured shadows as she is. The same can be said for her fellow agents since they’re all basically her with different faces. If textured shadows were necessary, then you’d see them mostly under her lower lip, eye lids, and nose. However, the latter two could be mesh shadows instead.

Sidonia no Kishi

TLDR: Poor facial animation, characters look alike, simple lighting, robotic animation.

To me, the animation of Sidonia no Kishi was a little disappointing when I saw the characters. All of the characters are clones of each other. No, I don’t mean canonically that they are clones (some are, especially those sisters pulling off the Nurse Joy effect), I mean that the models are exactly the same. While the main character Tanikaze Nagate has his own model as well as a few other characters, some of the main characters share face models with others. Shinatoze Izana shares her face model with a generic character from the ship. You won’t notice it since their hair styles are different and you never see their faces together. However, I actually downloaded the free models and bought some premium ones when they were selling them for study purposes. That’s how I found out those two characters have the exact same face. However, you’ll still notice this even if you don’t download the models. Basically, they have 3 faces for each gender, and 3 hairstyles to mix and match, essentially creating 9 characters that you don’t care about. These characters are easily seen on the street in either academy uniform or civilian garb. You can argue that it’s because they’re all clones, but since it never got addressed, it can easily be perceived as dragging and dropping generic characters.
The animation isn’t super amazing. Most of the characters have robotic movements: their heads turn at a constant rate, their eyebrows change to show emotion but their eyes don’t as much, and their bodies are too stiff. I would say that most of the animation went into the scenes where they’re fighting Gauna in their Gardes; at least the mechs moving like robots is understandable. You will see in season 2 when one guy gets riddled with “bullets” that he doesn’t look any different dead or alive. Ironically, dying is the most alive the character ever looked.
The facial animation in Sidonia no Kishi is pretty much the same as any other. However, they are not as expressive and don’t use those slapstick faces such as (O_O) or (>o<). They are consistent with the size of the mouth and will only change its size if it makes sense, such as yelling. They do the usual lips going up and down as the method of talking so everyone appears to be saying “wah wah wah wah” and could probably be easily replace with that voice clip form Charlie Brown. They’re a little different in the fact that they show the teeth more often, but they still do that thing where the upper teeth can freely move around the head because they’re rather not have the jaw open too much. Understandable, as that could ruin the shadows of the face, but for some people it just doesn’t look good in 3D.
Speaking of shadows, the animation seems to be more reliant on one source of lighting. This source of lighting sometimes changes depending on the angle of the next shot, but it is generally in the same area. This is most notably seen during lectures and briefings where the source of light is the screen with all of the information. Unfortunately, the characters suffer when the light is behind them and the face is all one color because the nose doesn’t have a line or any extra detail to show where the nose is. This is also problematic when the characters turn their head too much. The shadow of the face can look like a nice crescent at one point, but when the person looks to their left, then a weird shape starts to form. The lips also start to look strange, especially when the head tilts down since the shadow they cast makes the lips look huge, which are huge on the model.
The texturing of this show is a little more in-depth. While everything looks gritty and dirty, especially Nagate’s skin suit right after becoming a new pilot, it’s fitting for their situation. I’m not particularly skilled in backgrounds and environments, so I can’t talk about their texturing as much. However, the character’s faces show various textures throughout the series. Sweat drops, blushing, and fear are some of comedic examples. When they show anger, rage, shock, or physical exertion, the appropriate lines and shadows are added.

Ajin

TLDR: Good facial textures, facial animation needs improvement, characters move too much or too unrealistic, simple lighting.

It’s the same studio as the one that made Sidonia no Kishi. Initially it is different because they have (I believe) Africans (who speak fluent Japanese for some reason) in the opening scene. In my opinion, in both 2D and 3D, they tend to make non-Asians look more realistic, or at the very least more distinctive and prominent features. You can already see the extra details that they’ve added into the animation, such as the gunman’s fear when he is about to shoot the Ajin.
At least the characters here all look different. The ones that do look the same are usually uniformed officers. You’re not constantly bombarded with people’s profile pictures to see that they all share the same face, so I could be mistaken and they just did a better job at hiding it this time.
The animation itself has been changed since Sidonia no Kishi, although I can’t really say whether it is better or not. I’ve criticized Sidonia no Kishi for the characters acting too robotic, but in this one the characters can appear too animated at certain points. To me, at certain points, the characters move around when they shouldn’t, at least not to that extent. There’s one instance when the two garbage men get guns and the one on the left shows his excitement with strange body language. I could be mistaken since this is Japan and they have different body language. When the characters move around or talk, their bodies move somewhat, but when they stop talking, stop moving, or show no emotion at all, they proceed back to stiffness, and the transition to that doesn’t even look good. Around 11 minutes into episode 8 where Tosaki is talking to the guy in glasses, each character only moves when they are talking.
The facial animation is better, but there’s still a lot to be desired. The same Charlie Brown thing can be said for this series as well. To me, it doesn’t work out well in 3D. There’s definitely a wider range of emotions that these characters can show. Nervousness, laughter, anger, sadness, they’ve definitely added more animation into the eyes and mouth, although I can’t tell if they’re blendshapes/shapekeys or bones. Most likely bones as you can see Tanaka playing with his lips in the first episode. However, if they had this sort of animation built into the characters, then I am sure that they could make the mouth actually follow the dialogue. And no, I don’t mean for the scenes when they get up closer to the character’s face, I mean for the whole series.
The shadows are not different in this series. They are still using the one light source method. However, they’ve solved the problem with the face looking flat when covered in shadow by either making the nostrils have lines all the time, or by adding lines to the rest of the nose. While it’s not pretty, it works.
Finally, the texturing of this series has been improved over Sidonia no Kishi. Well, at least they use more textures on the faces. And no, I’m not talking about when they’re dirty or bloody or anything like that. I mean to help with their facial expressions. The best example of this is when Izumi Shimomura gets stabbed by an IBM and is hoisted up into the air. You can see that they added textures on the bridge of her nose, inner eyebrows, and right under her tear duct in order to show her pain. This can also be used when trying to show anger or squinting eyes. I wish that they added some shadow to the corner of her mouth or near her nose to make it more believable if people were just looking at her mouth. They used textures around the mouth more in Sidonia no Kishi, but in Ajin it seems like those textures are now just cast shadows with the lower lip being the only area with a textured shadow.

Aoki Hagane no Arpeggio

TLDR: Better facial animation, lighting and shadows seem manually changed constantly, character animation is decent.

The animation of this one is very interesting. Most of the stuff I’ll be talking about in here have to do with the movie since the animation improved in the movie. In the anime, you’ll definitely see robotic movements, such as when Gunzo turns around to talk to Iori on the screen, but he turns with his head and torso at nearly the same time and rate while his eyes stayed locked forward wherever his head is pointing.
Unfortunately, with this series the animation tries too hard to pull off 2D and they end up making everything look robotic. To me, the character that behaved the most different from everyone else had some of the best animation. Ashigara is one of those energetic types; her moving around/boxing during the Student Council Meeting (yes, there is a student council… for ships) and her actions during the battle looked very good for a 3D anime. She had some parts where she went stiff or didn’t transition into a still pose quite well, but that’s pretty much what all the other characters were having problems with as well. They’re not all terribly robotic, but still they can fix a few things. During the movie, Hiei does a turn around a 1:16:20 where the order of her movements is eyes, head, torso. While that part was good, her stance afterwards was not as much. She turns, stops, changes stance again, and then stops again. If you mute this scene, you’d think that someone else was talking to her and she was reacting to what they were saying. In fact, read this script while the scene plays on mute.
HIEI: I’m gonna kill you!
OTHER PERSON: Just try to do it b****.
HIEI: Nani?! (What?!) Don’t make me come up there!
Although if you watch it, you’d know that it was her talking the whole time.
The facial animations are much better. They pull of slapstick faces quite nicely. The lip-syncing looks like actual lip-syncing instead of Charlie Brown speech. Granted, there’s only a few scenes where they do it, but when they do show, they seem much easier on the eyes. When Takao and Ashigara are fighting with melee weapons, both of their facial animations match the words. Or when Musashi goes on her temper tantrum, this is where the lip-syncing worked perfectly and the Charlie Brown effect would have ruined it. Watch other anime to see if they do the lip-sync during moments of high emotion, or just yelling. The only example I can say is in episode 2 of Muv-Luv Alternative: Total Eclipse when the girl screams “Shoot me!” However, her mouth doesn’t follow the “Bu-Te-Yo!!!” her eyes don’t widen, and there’s no shadow around the mouth to show emphasis on her screaming. Mute it and see if you can really tell if she’s screaming in fear. Back to Arpeggio, it looks great, except they could have put more emphasis into her mouth being wide open as she screams in the form of a shadow, crease, or line. But I guess that wouldn’t look cute on a loli. I dunno.
They don’t do too much texturing in the face as far as I’m aware. If the lips were textures and not just different materials assigned to the mesh, then you can consider them in this. The lipstick and the shine on lips (lipstick or not) would probably be the most texture that they do on the faces. Most of the texture goes into the hair. If you study each character, there’s many inconsistencies with the hair. Most of the time, they aren’t affected by the light source. All of their shadows are either built into the texture map, or are a separate shadow map. At some point, the hair doesn’t cast over the body, especially from a distance.

A Butterfly’s Dream

TLDR: Too short to study every detail, great animation, hair simulation is good, much more appealing than above series and movies.

            I am really glad that I found this commercial on Vimeo. They’ve included a breakdown of a few scenes and I’ve studied them several times already. Since this is a commercial, there’s unfortunately not enough material to study as much as I’d like, such as lip-syncing.
            To start it all off, you’ve got the main character who really likes yellow, so let’s just call her Yellow for now since I don’t know her real name. Her hair has yellow strands and she uses yellow nail polish. You get two other characters, but you only see them running down a hallway and chatting around Yellow while having lunch on a bench. Any other character in the commercial are only seen from the back in lab coats while Yellow gives a presentation, assumingly about butterflies.
            The character animation for <i>A Butterfly’s Dream</i> is really well done. This is understandable since this is a commercial and they A) would want to advertise the school as best as they could since presentation is very important (notably in food at maid cafés were the appearance of the girls and the food matter more than the taste of the food), and B) are able to put more resources into a short project and make it look better than an animation studio trying to make a series out of 25 minute episodes. The walking/running animation is realistic, they added in extra gestures to make it look like they are really in a hurry or are just excited to get somewhere. This can be seen where Yellow starts off doing a skip backwards while telling her friends to hurry up. If you mute the scene, the same message comes through. The scene afterwards is great when they show something catching Yellow’s attention: She’s laughing towards her friend, only her eyes turn towards the camera, then her head and torso follow when her attention focuses. The scene in the library while she searches for a book is really well done. Her eyes are scanning the books, she looks at the next set before her head turns, then her eyes look up before her head tilts. Pretty much every scene that you see her in, she blinks, making her seem that much less robotic. Every other animation that I’ve listed above either has characters not blinking, or waiting too long between blinks. Granted, I know people can go for a long time without blinking, but it looks strange when every character lets their eyes dry out.
            Facial animation is great. They do a good job of positioning the mouth when the camera is looking from an angle like how it would if it were done in 2D. This is not seen during the scenes when the camera is looking at the character from a ¾ view and then she turns her head to face the camera since that would look weird. Not much to look at in lip-syncing since she doesn’t talk for an extensive amount of time. I guess when she says “Ki-re-i” (so pretty) could count. However, the small gesture in her mouth in combination with her eyes work really well to show off what she is trying to do in the scene. Speaking of her eyes, they are just a texture, not an actual mesh. You can see this in the breakdown, or whenever she turns her head too far in one direction. Since Yellow never turns her head to where one eye comes in/out of view, this is passable.
            When it comes to the backgrounds, texturing is good. However, I said earlier that environments aren’t my strong suit, so I’ll skip that. For the characters, there’s not too much texture going on with them that I haven’t mentioned already. Like I said, the eyes are a texture, clearly seen when Yellow is looking for books and the shadows on her face and eyes align perfectly. She blushes at least once in the story, so I guess that’s texturing, although I could be mistaken and they could be using something else for that.
            I studied this animation mostly for the shadows. They managed to maintain the shape of the shadow, but that’s because the character never had to turn her head too much, or the position of the camera and the light source were put in places where the deformation wouldn’t be seen. My best example would be around one minute into the video when she turns her head towards the direction of the light. You can see a VERY subtle change in the shape of the shadow, but not enough break the crescent shape. Even during scenes where most of her face is covered in shadow, she never turns her head away from the source of light during the animation. However, I could be mistaken and all of those shadows were actually shadow maps that were manually edited in Photoshop afterwards. My next paragraph will explain why.

Her hair is very interesting. During most of the animation, her hair waves around a little bit with probably less weight than hair would actually have. Whenever the camera goes up close to her face and it is a windy scene (0:19, 0:22, and 1:00 in the video), her hair isn’t actually the mesh anymore. If you watch the breakdown, you’ll see how they did the 0:19 scene. The mesh actually doesn’t move around that much, so they used the mesh’s simulation as reference, and then created a more turbulent hair movement in Photoshop. Then they manually created the shadow map, lines and highlight (I don’t know the name for that map) and put it all together. One thing that you may notice is that on the right side of Yellow’s face (from viewer’s perspective) is that they removed the yellow streak of hair in the back and just made it brown. They did it as well in the other mentioned scenes, so that’s what tipped me off to those being manually redone in Photoshop, along with the shots being so close to their face.
Of course, as I mentioned earlier, I could be wrong about a lot of this. All of the shadows on the face could have been a manually created shadow map. It’s certainly possible, and they’re just really good at making it seamless. Those other scenes may or may not have had their hair redone manually, or there could be more scenes where the hair was redone manually. Doubtful since I don’t think they’d worry too much about shots where Yellow is too further away from the camera. Still, that’s a lot of effort into this commercial, and the end product is excellent because of it.

Tldr…

This was quite a long read but had some interesting points. I have also been curious about 3d used for 2d anime workflow and how it could look more appealing graphically. I think what interests me most is the hybrid workflow or drawing over 3d to keep consistency and fluidity while keeping the 2d anime look.
Here are some examples:

Done by a talented artist-Mike Inel. He sometimes uses 3d models/environments as references for his 2d animations to get complex animations, character detailing and camera movements.

Another interesting video using 3d as reference for 2d animations.

A discussion about cg and anime.

I feel the right way to improve 2d animation especially with details is using 3d as reference. Though this might be very time consuming.

I also wanted to ask here. Between 2d and 3d which do you guys think is easier?
According to this vid;

The guy says 2d animation is easier than 3d. I think 3d is easier in certain situations especially when it comes to reusing assets and being dynamic with your projects. If you can be patient during the modelling, texturing and rigging phase, the animation part becomes a breeze. But then rendering all the frames if you are going for realistic graphics would take time.
2d most definitely is less complex and cheaper especially with the render time and no need for rigging, technical knowledge and creating assets.
I think using both, can give you better and faster results.