2 days to Vegas. Next gen game

[Farid]

I'll say... I don't know how they get reflections of a city from a grey background.

That's because they're using a ramdom HDRi background for the reflections on their models.
It looks indeed completely "OFF".

 

[ShootMyMonkey]

In all honesty, the depth and complexity of the lighting I see looks very reasonable and possible on next-gen hardware. The thing that throws me, though, is the model complexity. Sure, those cars have nice lighting, but do you notice the apparent polycount on them? Possible in a GT-type game where the primary focus is just the tracks and the cars. But in a GTA-type game, where you've got a much more expansive environment filled with lots of active characters... not likely.

 

[Laa-Yosh]

The lighting is far too complex.

The other giveaway is the quality of the AA and the texture filtering - we can probably expect 4x multisample AA and some anisotropic filtering from the next gen, but this looks a whole lot better. Thus it is the product of an offline renderer.

 

[_phil_]

do you notice the apparent polycount on them?

yep ,overkilling quantity,and the rest of the scenery around lacks,and the buildings are a copy/paste of the same set of 3,the progressive soft shadows,the dynamic ambiant occlusion or final gathering.everything shouts cgi rendering,and not even with realtime models.

 

[Shifty Geezer]

Are pics like these gonna hurt Next-gen at all? If they don't come up with the goods, aren't we all going to be a little disappointed?

I'm rather wishing they'd keep all the promo shots for the publishers, and not show the game until it's finished. That way it'll look good compared to current-gen, instead of bad compared to promo shots. I also feel the same about screenshots for games in the marketting lead-up during development. Actually post real pics, not fluffy AA'd glitz shots that the games hasn't a hope of ever producing.

 

[London Geezer]

Actually post real pics, not fluffy AA'd glitz shots that the games hasn't a hope of ever producing.

Heh, they've been doing it for years with current gen games, what makes you think they'll just stop and be good.

 

[Shifty Geezer]

Just 'coz it won't happen, doesn't stop me wishing. I also wish for world peace and a pony, but neither will happen

 

[Gollum]

In all honesty, the depth and complexity of the lighting I see looks very reasonable and possible on next-gen hardware. The thing that throws me, though, is the model complexity.

IMO exactly the opposite is the case. It's still easier to push around tens or hundreds of thousands of multi-layer-textured polygons in real-time (normalmapping helps to keep polycount at sane levels though), than it is to make expensive lighting and shading calculations, let alone radiosity.

The lighting in some of these shots is defnitely too sophisticated for real-time IMO. Of course a radiosity solution could be pre-baked as a texture- or vertexmap and there are some realtime radiosity solutions in development, but I doubt we'll see this level of quality from them within the next generation. The shading noise, especially in the car close-ups, is a dead giveaway for an offline renderer. Such noise is typical for Monte-Carlo radiosity and/or soft shadow raytracing and highly unlikely to be found in a real-time engine.

I actually think a next-gen game engine could achieve results similar to the city street shots with lots of clever faking and baking (and help of a realtime occlusion solution, if available at that time), but the lighting/shading quality would still be noticably inferior even under optimal circumstances IMO...

 

[ZoinKs!]

Are pics like these gonna hurt Next-gen at all? If they don't come up with the goods, aren't we all going to be a little disappointed?

Only the most hardcore of the hardcore will pay much attention now and remember these shots a year from now. You have to remember: no one on these boards is normal.

 

[London Geezer]

You have to remember: no one on these boards is normal.

HEY!!!