Metal Gear Solid 4 gameplay trailer @ TGS
- Thread starter one
- Start date
scooby_dooby
Legend
That's completely beside the point. You're making this personal which it doesn't need to be.
there's nothing 'personal' here - it's a discussion between Laa-Yosh and me. over graphics. on a graphics-oriented board. if you don't like my use of the word 'picking', i can assure you i don't imply anything inherently evil with it.
That low? It's painted by hand. Into the texture. And not just on Snake but on every other character, Gear, and weapon.
Or, if you disagree, provide a reasonable counter-argument, other than 'I find it unlikely'
I'm having trouble to see how this works - Snake's shading looks different in every angle and setting that I can find. I do see that there is banding, where you have pointed out, partly highlighted from those shots having high contrast ratios, that are decreased in later shorts. Regardless of what Kojima will eventually settle on, they do seem to point out a weakness in the shading. But with Snake's face being so well animated, I'd think that painting that kind of stuff by hand would be a lot more work than actually calculating most of it.
We're both guessing here, not being able to look at the textures, shaders or light setups.
And there's no factual evidence to support the opinion some of us have, that there's been some grapchical downgrades since the first techdemo, either.
So, I think we can go on with this theoretical discussion without offending each other... right?
My problem with the banding is that they'd need to use 8-bit per color shaders to get such effects (remember the original Halo?) and even then it's very very extreme.
Shading painted into the texture... take a look at the clothes of this guy:
http://images.playsyde.com/gallery/public/4361/899_0013.jpg
Shirt and pants clearly have it. Snake's suit is very different so the effect may not be this evident.
http://images.playsyde.com/gallery/public/4361/899_0008.jpg
http://images.playsyde.com/gallery/public/4361/899_0002.jpg
Compare the bolts or whatever on the suit near the lower end of his delts and the area right above his elbow. The colors look almost exactly the same, yet there seems to be some difference in the main light's position.
Please note: I don't want to say that they've completely dropped any dynamic lights.
What I think might be going on is that they have dynamic lights and normal mapping and speculars too going on in there, but they've found that more than one light source is too taxing for what they're doing, or that they'd need too many light sources to get the look that they want (remember Doom 3?).
So they go with 1-2 lights and constant ambient light for the areas that don't get any direct light. This would make normal mapped models completely flat so they decided to paint some shading into the diffuse texture. There's nothing bad about it, many people do it even in offline CG.
Again, the point is that I'm trying to find out what the difference might be between the announcement techdemo material and these new images. Some people do think that there's a difference, and so far I haven't really seen any other explanations other than 'different art direction'.
No one would really notice the difference between using an occlusion map and a diffuse texture with the occlusion baked into it.
Not that there's any strict science for occlusion as it is a big big hack. We can use the occlusion pass to multiply the enviroment light pass only, or we can also multiply the direct light passes too. Whatever the art director prefers is what matters.
Also, on animated objects, what you can bake won't account for moving close to a wall that occludes one side of the character, or the effect that the lowered arm has on the sides of the body, or the gun held close to his chest... which is also very important, it's not just about the high frequency details.
But I don't see any pre-baked occlusion effect on MGS4 characters...
Only color variations for dirt, wear and tear, some randomness to make the material less artificial. Note that areas more exposed to the outside world are more likely to fade out, get the paint off, suffer damage etc. etc., whereas dirt accumulates in holes, skratches and other low points. But we don't paint diffuse shading into the textures... I can't show you any texture images though, sorry.
(Note that we do CGI, not games... but if you browse Doom3 textures, most of them don't have any diffuse shading either. I'd love to see Gears' textures, though...)
And there's no factual evidence to support the opinion some of us have, that there's been some grapchical downgrades since the first techdemo, either.
So, I think we can go on with this theoretical discussion without offending each other... right?
My problem with the banding is that they'd need to use 8-bit per color shaders to get such effects (remember the original Halo?) and even then it's very very extreme.
Shading painted into the texture... take a look at the clothes of this guy:
http://images.playsyde.com/gallery/public/4361/899_0013.jpg
Shirt and pants clearly have it. Snake's suit is very different so the effect may not be this evident.
http://images.playsyde.com/gallery/public/4361/899_0008.jpg
http://images.playsyde.com/gallery/public/4361/899_0002.jpg
Compare the bolts or whatever on the suit near the lower end of his delts and the area right above his elbow. The colors look almost exactly the same, yet there seems to be some difference in the main light's position.
Please note: I don't want to say that they've completely dropped any dynamic lights.
What I think might be going on is that they have dynamic lights and normal mapping and speculars too going on in there, but they've found that more than one light source is too taxing for what they're doing, or that they'd need too many light sources to get the look that they want (remember Doom 3?).
So they go with 1-2 lights and constant ambient light for the areas that don't get any direct light. This would make normal mapped models completely flat so they decided to paint some shading into the diffuse texture. There's nothing bad about it, many people do it even in offline CG.
Again, the point is that I'm trying to find out what the difference might be between the announcement techdemo material and these new images. Some people do think that there's a difference, and so far I haven't really seen any other explanations other than 'different art direction'.
No one would really notice the difference between using an occlusion map and a diffuse texture with the occlusion baked into it.
Not that there's any strict science for occlusion as it is a big big hack. We can use the occlusion pass to multiply the enviroment light pass only, or we can also multiply the direct light passes too. Whatever the art director prefers is what matters.
Also, on animated objects, what you can bake won't account for moving close to a wall that occludes one side of the character, or the effect that the lowered arm has on the sides of the body, or the gun held close to his chest... which is also very important, it's not just about the high frequency details.
But I don't see any pre-baked occlusion effect on MGS4 characters...
Only color variations for dirt, wear and tear, some randomness to make the material less artificial. Note that areas more exposed to the outside world are more likely to fade out, get the paint off, suffer damage etc. etc., whereas dirt accumulates in holes, skratches and other low points. But we don't paint diffuse shading into the textures... I can't show you any texture images though, sorry.
(Note that we do CGI, not games... but if you browse Doom3 textures, most of them don't have any diffuse shading either. I'd love to see Gears' textures, though...)
by all means. i see that now we both agree on that.
did not mean to be offensive and i hope you did not take anything i said this way.
yes, the bolts are potential candidates for pre-shading. still, take a look at these two shots - A and B - the bolt on the back-side of the right-hand delta: on A it's rather uniformly lit, but still slightly shaded at 2 o'clock and highlit at 8 o'clock; on B the shading is reversed and is way more prominent - dark at 9 o'clock and lit at 3 o'clock.
overall, in no shot so far do those bolts strike me as depicting light from an apparently wrong direction. if any, their problem might lie in the resolution of the normal map spanning them, at least the ones around the back/shoulders/elbows.
yes, from my experience i'm fairly sure they have intensity variations somewhere in their diffuse maps (hell, if nothing else they have some radiosity diffuse maps on their environments). but the question is where/when. re snake's model in particular, it does seem to be fairly highly light-teractive so far. btw, today i went carefully through one of the full-res trailers kojima released earlier this year (the one with the arabs convoy ambushed by mercs, eventually wiped out by bots) and i noticed something interesting there - in the open, in a bunch of situations where i thought i had spotted potential diffuse pre-bakes, the object would eventually re-orient versus the light direction and those shadings would turn out as being fully dynamic. except in one place where the shading was very subtle. as usual, proof of dynamic lighting is best descerned in dynamic footage.
Last edited by a moderator:
Its not the game, its the footage. All the footage is way to bright due to useing a camera and it makes the colours look washed out, here are some proper screen's
The crappy footage does this game no favors at all.
Mr Boo's pics look very similar to the original Snake/Otokon demo. The same fancy lighting is there and it doesn't look as flat as some are trying to suggest. Remember that Kojima said the footage was all running realtime on PS3
I beleive the 'direct capture' screenshots don't really represent what the game will look like on PS3. They have an extreme amount of AA which proves it wasnt taken from a PS3 running the game in real time. Konami might have been running the engine on a PC with a lot of effects cut out (just a guess).
So what Im trying to say is that Mr Boo's shots are a truer representation of what the game will really look like on your TV.
Eagle-Vision
Regular
I don't know, small and dark screens tend to hide detail as well as flaws. For texture resolution, lighting (well, not dynamic stuff obviously) and over image quality (other than the AA) I think screenshots are very reliable.
They do tend to hide details but look how much colour the pictures i posted contain compared to the official grabs. The grabs make the game seem really washed out and dull, were as going by my picture's the game is much more colourful then the grabs imply.
mckmas8808
Legend
Of course. You're right and that's what I've been trying to say through the last few pages of this thread.
Eagle-Vision
Regular
It seems to have more to do with the higher contrast, which is probably caused by the camera, not the actual footage.
mckmas8808
Legend
No the colors actually do look like that. Every single video I've seen has the colors that look like that. And it goes inline with the same colors that we've seen from direct capture video earlier this year.
scooby_dooby
Legend
So the video looks liek that, but when you capture the frames directly, they somehow lose all colour and contrast? That's impossible, a frame is a frame, the video is nothing more than a series of images.
mckmas8808
Legend
What are you talking about? I'm agreeing with bleon that the video represents the colors better than the screenshots do.
What pictures or videos are you looking at?
Eagle-Vision
Regular
The direct feed screenshots have the same colors. It's just these screencaps are darker which makes the lighting seem "less flat".
Last edited by a moderator:
scooby_dooby
Legend
How can images taken directly from the video lose colour?
Because the video colour space and image colour space are not the same. This was a problem about a fortnight ago with some of the TGS screens appearing way too bright and washed out, much like these MGS ones do. It turned out then the capture device was reading a source it thought was at contrast/brightness X and converting it to Y when it was a source with contrast/brightness Y. I can't remember the game but if you search I'm sure you'll find it in this very forum.
scooby_dooby
Legend
Taking an accurate screengrab from a video file is not exactly rocket science.
