THE iDOLM@STER > THE iDOLM@STER 2

iDOLM@STER 2!!

<< < (146/608) > >>

Cael K.:
Second part, up to 10 minutes, when this segment ends and the animation segment begins.

0:20 - Toon shaded character against a non-toon shaded background (first pic, second top left). It was kinda jarring, so they applied soft focus (top right), and messed with the depth of field (bottom left). The bottom right shows both effects in place. They're similar effects (both are blurring effects), but they do slightly different things.

2:30 - They're using light bloom for a more HDR feel. Check the lights in the background. They also say "flare," which I assume is lens flare, but he doesn't mention it.

3:40 - Oooh, totally missed something here, but for the middle camera (and the long camera), I think they were saying that the depth of field wasn't really working out as well as they'd hoped.

4:20 - No matter how I look at this, the entire thing looks like an action pic. Yayoi totally parried the Yukiho Chop, Takane is looking like she owns the world. Let's get back to the main topic, though... they're using light bloom, depth of field, soft focus, and lens flare. Everything's rendered using 4x multi-sampling anti-aliasing, which pretty much means that the source is rendered at 4x resolution (so 1900x1080 becomes like... 7600x4320), then downsampled to the destination resolution. This is done to reduce jaggies or visible seams where polygons meet.

5:20 - The shadows are not consistent here.　I think I know what he's getting at in the second picture, though. Self-shadows are done for objects individually, and the light source used for one idol's self-shadows isn't the same as the light source used for another's. This is why the shadows in the first picture aren't consistent with each other, since they were calculated independently. However, they can't go to using a global light source for everyone, since that might cause shadows they don't want.

7:50 - Clockwise, from top left: new shader, post-processing, self-shadowing, five idols, 4x multi-sampling anti-aliasing, 60 fps. The caption: To reduce increasing load, all these systems must be redone... and he adds the fact that they scrapped them completely. As an aside, the details will not be in next month's Gemaga, even if they say they're revealing secrets.

9:00 - Pretty much everything changed. I think he says they stopped using the middleware they were using, and data formats for models and such changed around. Some people are saying this is why the game isn't coming out on the PS3, which I think is... not entirely right.

For those of you who want to find out more... he intends to talk more about this at the Tokyo Game Show. This makes me sad, as I can't go.

Next is the animation part.

Cael K.:
Stuff from the animation chapter.

12:00 - He's going to show braids, long hair, and how to make things not clip into each other.

13:30 - Not doing anything to the braid, it's stiff like wax were applied to it. Pretty easy to see.

14:20 - From top: Make the model and skin it [translator: I guess this means texture it]. Move the joints placed in the hair. (Green) Position before [translator: at -1 frame]. (Blue) Speed, gravity, forces which cause the joints to move back to their original position, etc. (Green) Position after [translator: current frame]. All this just means: every joint in Yayoi's braid here has some amount of tendency to go back to its original position, and is acted on by various forces like their current momentum and gravity (the returning forces mentioned before is actually one of these). Combine all these forces, and you've got your new position for one of these joints. These joints also have other properties to them, like facing (for twisting the hair), but these are ignored for now.

15:50 - (Top) Speed times some resistance (attenuation) factor. (Top green) The joints move by inertia, but speed eventually slows down naturally [literally: speed decays]. (Bottom green) Apply some resistance factor to speed when calculating joint position. Naturally, there must be some constraints to this, since if a joint moves on its own, it might move further away from the next (or previous) joint in the braid, then the braid becomes longer.

16:50 - (Top) The length between joints cannot change, and the base of the braid cannot move. (Bottom) The entire braid is pulled by the movement of the base.

17:50 ~ 20:10 - So, just applying speed and its decay... well, you get it. They put gravity and wind in, and... the feeling of volume is missing. So, they put in some more joints. These joints pull on the braid, though they don't actually fall on the braid (in fact, they're just points in space), and pull it back to its original position. With everything in place, Yayoi's twin tails animate quite well.

21:30 - Just a review of all the points above. (Top) Speed and speed abatement, gravity and wind, braids returning to their original shape. After all these are done, the length between joints is always adjusted to its original length. (Bottom) The important variables are: Speed decay factor, gravity, and the force factor for returning the hair to its original shape. (Bottom right) Tails all done!

22:00 ~ 24:20 - Moving on to long hair! First, hair has many lines of joints. The first movie shown has less than the second, I think. They put have spring forces between joints, to keep them from flying apart from each other. Essentially, you can think of them as a spring: if the connected points get too far apart (here, look at the red lines), they pull each other together, and if they get to close, they push each other apart.

26:10 - (Green) Long hair's done! (Bottom) Even if we're calling it done just by how it moves...

26:30 - Collision with swaying objects. (Top) Hair and skirts bump against the body and legs. (Middle left) Swaying objects use spheres. (Middle right) The body uses cylinders and spheres. For collision detection, a triangle-by-triangle test is incredibly ineffective, so usually they logically bound (encase) sections of the body with a combination of near-form-fitting primitives... simple things like spheres, cylinders, or boxes. These are the objects used for collision, not the actual triangles on the model, which makes it much faster. Note Mami's neck: there's a sphere here containing empty air, but regardless, if something hits the edge of that air, then it's considered to have hit Mami, in which case it bounces off or Mami goes flying.

27:20 - A list of all the types of objects used in collision detection. (Top) A plane, for floors and such. (Middle) Donut ring, for skirt edges and such. (Bottom) Tubes, for tail-like hair and such. Notice each of these is a bunch of objects roughly in the aforementioned shape (a tube can be a combination of spheres and cylinders).

27:40 - From top: How are spheres and cylinders used for collision. Cylinders are used for the body, arms, and legs. Spheres are used for joints. Swaying objects are usually just spheres. He shows these bounding volumes starting from the cylinders, going to the joints, then adding in all the swaying stuff.

29:10 - They couldn't show Iori for some reason, so we get Haruka and Takane. Notice the big cylinder behind Takane: this is there to keep her hair back.

Final slides.
31:20 - (Top) More fluid behavior (in animation).
31:25 - (Top) Additions and removals in flexibility. (Middle) Accessories didn't move around in L4U! (Bottom) Accessories sway around properly in Im@s2.
31:45 - Faster.
32:20 - From top: How swaying objects are processed in Im@s2. Simple movements, simple collisions, still with some great visual expression.

... And that's it. That took quite a bit. Personal opinions to come.

And the programmer, Maezawa Keiichi, would like everyone to remember Sensitive Tooning.

airgeneralpong:
Looks complicate and nice, I like it. Do feel good watching iM@S presentation.
Thanks for the vid and summary. q^_^p

Cael K.:

--- Quote from: airgeneralpong on September 04, 2010, 06:16:36 pm ---Looks complicate and nice, I like it. Do feel good watching iM@S presentation.
Thanks for the vid and summary. q^_^p

--- End quote ---

Thanks! Worked pretty hard on that. ^^; As far as my own opinions...

Okay, well I honestly thought the animation part was pretty standard. What I thought was interesting is that they put in another set of joints so that braids could snap back to their original position, since I thought you could do that without them. Thinking about it, though, it's probably easier (much easier, thinking about it carefully) to do it with another set of joints. I just gained Exp.

As far as shadowing goes, well... watching the PVs again, there are some times where they don't look consistent either. Also interesting to note: necklaces and such (even small chains on costumes) can sway violently, and they do cast shadows on the idol wearing them... but idols don't cast shadows on themselves. Maybe my eyes were mistaken, but it almost looked as if things like the lapels on costumes and maybe... the fur on Mami's costume moved a bit. The skirts do flutter around maybe a bit loosely, though I found a clipping error somewhere. Also, I'm not sure Azusa Ahoge physics have been implemented. Maybe it just defies physics.

With all that, though... shadowing on the ground. We can all see the little 4-way mini-shadow going on and think... why aren't they better? This is just a theory, but if we could actually see them come from a specific direction, then it might jar with the self-shadows we see (the lighting's done separately from stage lighting). If they were to stick a shadow on ground that coincided with an idol's self-shadow, then switched cameras, the ground shadow's position would abruptly change (remember that self-shadows are dependent on where the camera is). We can just fudge everything by making sure there's a shadow in about every general direction, but then... we get those cheap looking shadows. Even worse, if we stick too many of them on the ground, things look bad.

From here on out, it's speculation time, so whatever I say might be really off (though the same applies to the above). The original Im@s ran on the System 246 board, which was an extension of the PS2 board, from what I understand. I'm going to hazard a guess, then, and say Im@s L4U!, and Im@s360 did not use anything HDR related, since I don't think the PS2 was capable of it. Now, we're seeing some effects using HDR, so there might be more shininess in the new version. Either way, the amount of new stuff they're doing in Im@s2 might be a direct consequence of finally being free of the old System 246 board. The 360 was always capable of this stuff, but if it wasn't in the original Im@s, then I'm betting it's because it was hard to fit new stuff on the old stuff.

The comments about it not going to PS3, well, I can see why it's not at the moment since I think some stuff might be XBox360 specific, but weirder things have happened with ports (see Bayonetta, except give a year to port). I'm not too experienced with these things, but there were some comments about RAM being an issue, but I think the PS3 has the same amount of RAM the XBox360 does (both graphics processors can access 512 MB, the PS3 is slower with it, though). Once again: no reason Im@s2 can't come out for PS3, it can definitely handle whatever Im@s2 can throw at it.

TweenDoriru:
What Yayoi has are considered "braids"?

(also, did anyone else think she looked cute with the pigtails in that position where they hang lower than normal? :3)

Thanks for the summary :) It's interesting to see in detail what's going on (since I watched it prior to this). I think even the smallest details they changed and added really shine (at least, via the PVs we've seen thus far).

Navigation

[0] Message Index

[#] Next page

[*] Previous page