"Yes, but how many polygons?" An artist blog entry with interesting numbers

They should use Humus's technique

 

http://blogs.amd.com/play/tressfx/

Real-Time "strand" based hair in a production game, on GCN no less. Using GPU Compute. Although I've seen no metric as to the performance cost on say a 7850 -7870.

Isn't PS 4 supposed to be "faster" at compute though, or is it just more adept at mixed work loads??

Anyways I just thought it timely, I suppose one could argue that sence Tomb Raider is basically a current generation (console) game they had cycles to spare on PC.

 

speaking of hair, AMD is introducing new hair effects support with the PC version of the new tombraider.
have a look.


http://blogs.amd.com/play/tressfx/

 

tressFX

That looks nice, but, honestly, I think that by now this should be standard hair graphics.

I'm amazed to see cardboard/plastic hair at this point of the generation, even in "ultra advanced" PC games. I think more than a technical difficulty, we have a development issue, and this is proof for my point: something like this is technically possible with current technology. I just think that developers don't focus on hair too much these days.

Heck, even Aquaman Battle for Atlantis had cool hair for its time!

 

I say let's wait how it looks and performs during actual gameplay first. That said, I also find this type of visual development very exciting.

 

Well, YT video for tressfx in motion (Tomb Raider):
http://www.youtube.com/watch?v=oAX_veVl40A

With all due respect to AMD developers, I don't like it. It's pretty on screenshots, but I was expecting something better than that: way too bouncy flocks of 'curtain' hair, with a not very refined collision detection. Water doesn't seem to have any (noticeable?) effect on hair physics either.

Again, at this point of the generation I think we should see far better hair simulation. Even on PS1 days we saw some pretty decent (bearing in mind the technical limitations and polygon budget at that time) hair and cloth simulation; I remember Street Fighter EX 2, for instance, where hair and skirts were pretty cool, and now I'm seeing the same thing with just more polygons... Is this it, really?

Maybe I'm just feeling too bitchy, today...

 

looks like a hit and miss to me; in some occasions it looks decent when the hair falls back into place. but it also look too sensitive for the action gameplay.....a little too distractive.

I got the PC version today and will give my opinion later on to say if it's worth it, or if it makes a big difference.

 

Well, in the case of that video, I just think that the hair movement is just too much. It would probably look better with more subtle movement.

 

Would probably look much better on characters with shorter hair too. Waving around like strings in that video.

 

This is imo an example of technology just for technologies sake. It does not help to immerse the player, as it just looks ridiculous at places and way overdone. But I guess when they tweaked it to subtle, more realistic movements, they realized that for such subtle effects no one is willing to accept the (rather massive according to early tests) performance hit.

 

I played a few hrs of it on my PC, and the hair came out like glitch fest in action. her hair goes sporadic many times as if it went went through shock theory or something. her hair glitches up when pausing the game.

it looks pretty good though when she isn't moving too much. oh, and tessellation is barely noticeable on Lara; it seems to be more effective for the title waves in the game. will post some screen on the details when i get the chance.

 

made some pics of the hair in the new tombraider.

TressFX

Regular



TressFX hair physics is it better?

...... just barely, it can be unnecessary to use the option; you'll find lara's hair in a dimension of it's own moving based on it's own inertia...and at times floating over her shoulders. I found myself turning the setting on and off just just to enjoy the gameplay.

does it look nice?

occasionally, when action kicks up in sequences you might find her hair collapsing and stretching with way too much sensitivity. however, when the settings are not too action oriented you can take the time to admire her hair bangs moving around.




my last two cents.

compared to her standard hair (which has plenty of detail enough to enjoy) moves a little stiff at times; but hardly stands out in a way that detracts your attention when you plan to just play the game. TressFX is a feature that you may grow to like but may also find yourself wishing it moved as nice as it looks.

 

See now how far secondary dynamics on characters have to go? Told you all

 

That is hardly representative of how hair rendering is definitely going to be like for the next few years. Its like pointing at AMD's MLAA, with all its faults, and claiming that AA can't be done better. Its early days yet, lets see how the generation pans out.

 

What I am trying to say here is that this is a field that has seen very little development so far and it will require a lot of attention in general. It is very likely that new technology has to be invented as the approaches used in offline CG aren't easy or perhaps even impossible to translate into realtime.

Same for clothing and other secondary dynamics.

This AMD tech is just something like a first baby step and my comment was meant to be about the entire issue.

 

Oh I see.

 

The underlying physics are completely wrong. The hair direction is perpendicular to the back of the head, the hair extends along that vector, and then it's 'dropped' onto the shoulders/ So a pony tail lying over the left shoulder instantly flies over to the right shoulder when the head is turned, rather than resting on the left shoulder until a suitable force lifts it over. As a solution its completely fake.

However, keeping this on topic (), what's the polygon difference in TressFX? Looks low-poly-count to me. I think there's plenty of scope to improve hair-tech by increasing counts for hair.

 

I agree. As I said in my previous post, this doesn't impress me much: not very good physics wise, nor polygon wise. Well, I think the worst here are physics; with the same polygon budget I think that we could achieve better results, if physics were convincing.

Again, I don't want to believe that if we don't have better hair solutions it's just because of current hardware. Again! We had similar physics even in PS1 era! Why so little improvement, if current hardware is X times more powerful?

 

As you and Shifty seem to be concluding, the issue seems to be the need of refining the codework and software efficiency than the lack of raw power. The polygons needed doesn't seem to be enough to strain current-gen consoles.

 

I wouldn't think those are some pretty simple vertex shaders running on those polygons... The number of triangles does count a lot more there.