Graphical effects rarely seen this gen that you expect/hope become standard next-gen

This thread is great. I've got a few thoughts that I hope people on here might be able to answer.

In shooters, when will it be possible to see your shots register visually on enemies (in terms of lasting physical damage i.e ripping through armour, or taking chunks of flesh off an enemy?

For example, when playing Gears 2 horde the other day, I think it seems odd there is no visual difference between an enemy (e.g a boomer) which is one shot away from being killed to one with full health. I realize masses of blood squirts out when you hit them- but it's not lasting visual damage on the model.

Also, details like permanently damaging a body part (causing a limp for example)- I’ve seen it done in games like GTA 4, but it seems like it's a quick animation and then they just seem to ‘reset’ and carry on .
Is it just time restrictions, or hardware limitations that stop developers from currently being able to do this? I think these kinds of detail would add a lot of realism – even with the current level of graphics.

 

Valve did some neat stuff with L4D2, where you could blow off or damage specific parts of the zombies. At GDC they just gave a presentation of their tech...nothing fancy, but it seems to have worked well enough for them. The PDF is here.

 

Ever heard of GHOUL engine used in SoF 1, 2 and Jedi Outcast at least?
It did that in the year 2000

 

Valve has implemented creases in their approximate subdivision surfaces for the source engine. You can find the paper by a bunch of researchers at NVIDIA, Valve and Industrial Light and Magic here. The part I'm talking about is on page 66, but the entire paper is a great primer on subdivision surfaces in the context of games. They don't seem to make a perfect fit, but apparently Valve is embracing them and working them into their tool chain.

I'd be curious to hear what Laa-Yosh might think of this.

 

I don't mind the missing gore/violence, as in SoF or L4D, but I'd rather see the way they made the bullet impacts impact on the npcs in Killzone 2. In most games, it is either a canned animation or the enemies just fall down as a ragdoll, but both of that looks mostly bad.

 

Well what you had in SOF & L4D is decapitations/Limbs flying off.
I'd rather want a KZ2 type effect where bullets that hit actually wound the person & the wound is a visible bullet hole and the death animation in KZ2 are probably the best I've ever seen in a video game till date.

 

In SoF (or SoF2? can't remember if it was in both) the system was rather crude, but great considering when the games were released, but it did include "bullet holes", if you hit for example someone in the face, you'd get bullethole on face (if you can call it that) and back of the head blown off

 

I want 2 things:

1. Proper lighting
2. Destructable environments based on PHYSICS. Scripted destruction as in BF sucks ass, once you blow one house to pieces you have seen them all.

 

Thanks for the link MJP, I wasn't aware of that.

I'm surprised that in a heavy gore styled game like Gears 2, they haven't included a similar feature to visually damage specfic body parts. It does look a little ridiculous when you've unloaded multiple clips into an enemy and they just look exactly the same!

Is it the current hardware limitations which stop developers adding these kinds of detail more?

 

Art and time. Also what their goals were. It's not like Gears 2 was skimping on art in the first place.

 

Yes, I mean think of all the hours they must've spent [strike]painting everything grey and brown[/strike] creating a cool new post-apocalyptic world from scratch.

 

Haha. Geez if it is down to lack of time thats abit disappointing - I hope they make time to include it more next gen.

Afterall, firing at stuff is a pretty big part of shooter games!

 

Gears 2 has gibbing, they've invested some resources into that - tearing characters apart and all.

The problem with displaying damaged states on enemies - not to mention differentiate between multiple versions - in a fast paced shooter is that you have to seriously overdo it to register and even then you can be happy if maybe 50% of the entire audience is ever going to notice it. It's just not worth the time usually - blowing the enemies apart is usually more of a spectacle and a better investment.

 

Selective 4xAA on sub-pixel stuff, MLAA on the rest of the image.

 

That would essentially be end of "Teh Jaggies"

 

Not for shader aliasing though.

 

Selective supersampling

Humus to the rescue

http://www.humus.name/index.php?page=3D&ID=64

1 sample


3 samples


11 samples

 

Not so sure it'd be affordable in a real game; depends I guess. For something like God of War 3 or even Gears 1&2, it could end up being rather expensive (in the same vein as supersampling alpha test can be expensive). Does Just Cause 2 have this implemented? Also curious how this will affect specular on edges.



I wonder if it's any better or worse with deferred renderers that specify specularity in a render target?

 

These are all questions you'd have to ask the Humus himself.

Note that I had dynamic branching disabled in my screenshots since it doesn't seem to work properly on my GTX260 (makes the bumpmap look... sparkly).

 

mm.... I tried this on my 5770 at 1080p. 3 samples seemed to be good enough for me. At least, the bobbing camera produced shimmering that didn't look much better than 11. What is rather poor is putting the camera up in the air and looking down with the floor tile in the entire screen. The supersampling is affecting pretty much everything and lowering the framerate (for me... roughly 25-50% hit from 3-11 samples versus none), but for little to no benefit since the specular aliasing isn't noticeable at that angle. Of course, it'll depend, but... I guess what I'm trying to say is that the perf can be rather unwieldy!

I'd be curious to see how it handles Doom 3, now that I think of it.. (or Tenebrae!!)