Minggu, 03 Agustus 2014

Guilty Gear, Guilty Design: In Flesh, Bloods and Bones.

Recently E3 showed Guilty Gear Xrd demos. Fucking exciting.
I am not big into fighting games, but damn those techies.

For you who are not clear about what i am talking about.
Here's some cuts from youtube and it's simply not enough...because it's over 9000!

Guilty Gear Xrd -SIGN- Arcade Version Opening

It may seem like casual anime opening, handdrawn on paper and all, but it is seriously all 3D.
And that feat is really tough, given some consideration such as lighting, shading style that is somewhat symbolic than realistic.




There are many instances of the animation that combines with 2D drawing so it may look somewhat show-like, flat and dynamic in storytelling, but for most of the part, it still belong to its own realm of 2.5 Animation.
And it is very beautiful. Here's one from May's entire cutscene. Not spoiler free.


Guilty Gear Xrd -SIGN- May Story Cutscene

The whole game had been a conversation over forums such as polycount.
http://www.polycount.com/forum/showthread.php?t=121144

Here's one a direct translation from the japanese folks.
I really need to get a friend to teach me japanese language.
http://www.4gamer.net/games/216/G021678/20140714079/


But recently i found a nice post from someone's twitter.
It's from Panupat Chong's blog.
http://panupat3d.blogspot.com/2014/07/guilty-gear-xrd-part-1-model-shading.html
http://panupat3d.blogspot.com/2014/07/guilty-gear-xrd-part-2-camera-animation.html

He does a very nice job of translating most of the content from various sources into a technical summary, easy enough to understand for english reader too (after translation).
I am just abit too greedy to let this slip.
This post will be separated into two parts for easier writing.

First part here will be explaining overall building of rigs, model and texturing.
And before we get too far, views and translation here are based on my personal choice of words,
so you are free to correct me and check the previous links to give me more insight.



Before the original conception, the team went through a trial design featuring Sol and his replicator.
The idea is not to make a game but basically an art style test. The whole look is pretty good but i can't tell much when the shading style comes in motion.


For most of the part, the modelling is full 3D, anatomically correct. I had thought most of the styling comes from painting...but apparently the edge loop is anatomically correct for the very least.
The design were done in both Softimage and UDK, synchronized so that the final result can be seen in both engine and creation kit.

It is safe to say from here that the design philosophy is still accessible to anybody who studies from life. Of course much to the argument that anime is not "realistic".




The rig itself however consists of at leasts 170 points for the face, everything designed so that the animation can be fully controlled, including cloth and armaments. Everything handmade and tedious.
I won't even be touching about the team, as the rumour has it that it consists mostly programmers.
It has however, went through unsatisfactory tryouts and experiments throughout the decades and you can confirm that from various franchises such as Street Fighter or Guilty Gear Overture.

The fight scene in my opinion, however, is somewhat simplistic. Almost a waste considering the hardwork put into the modelling and all those animation rigs... Only during special cutscene does the design culture stands out, perspective switch and all that.


Guilty Gear Xrd -SIGN- Ramlethal vs Bedman

The background also went through rigorous work, but more mainstream as it only follows the classic paintover 3D. It's a breeze to watch the wireframe though.



Traditional Cell Shading for animation is considered old standard and simple. But through examples from games like XIII (ubisoft) and Gravity Rush, Guilty Gear generates something new.





As said, it's basically generating another black object that serves as a silhouette to the subject using abit of z information from the camera, it'll bleed around the objects. I am not quite sure if toon shader and outline in maya is an equal replacement, but judging from the game's examples... the silhouette stood as its own unit in the gameplay.




From here on it's basically abit too technical for me, and my inability to read japanese does not help.
The engine basically avoids using traditional normal calculation for the shadowing method... instead vouch to read and write the shadow system from simpler forms.

Basically by manipulating the normals in various area, or using a formula called Gator(Generalized Attribute Transfer Operator) the system recognizes how shadow would be perceived in general sense within the games' own universe, and creates an overall image as according to target shape for the normal transfer.


Note that this Gator is not GatorCAD.... maybe.
Below is the illustration of Gator's target shape. It ignores the overall shape and conform to the target source. Some adjustments are made too, like tiny details and rendering layer, but in general...i don't read too much to those.





There are also the case of linework and the silhouette, which uses some camera tricks to fool the eye and make the overall style more consistent. This is a very tiny case but the major problem comes when there are more parts to be emphasized that is not the edge of the object.

Note how the hair is less detailed by the end but the overall information reads much better.
I am still not convinced though how such subtle details is an emphasis when the final product features so many still frames. But i guess quality control is number one.

And also the general problem of resolution when using textures to read out shape design. Studies how "jaggies to manifest in texture mapping out what that situation" and, whether issued through the how the "clean line segment that does not depend on the resolon of the texture."


Overall I'd like to point out older games such as Naruto and Bleach that has already tried various similar approach to the subject by making the 3D modelling conforms to the texture in a way that is similar to 2D vector art, and then baking the general form right away to the texture... also not forgetting the silhouette model as mentioned above to form the edges.
Bleach Versus Crusade All Cinematic Bankai

Although unfortunately it fails to address overlapping shadows with its texture maps and its real time frame rate that somewhat breaks the illusion too much, either by show jagged image or by having too much motion. But that last one will be something i touch upon the next blog post about Guilty Gear XrD's animation methodology and this new pseudo-handdrawn 3D design they just broke through.

Tidak ada komentar:

Posting Komentar