[Patch] Specular Highlight

[HarryTuttle]

I found an image of Fighting Vipers 2 from a scanned image of a magazine

Good catch!

Here are some video captures from VHS footage of the real arcade version (the best I could find). In particular, from here you can watch Grace in that stage, albeit the player lost there, so there's no winning pose. But look to the subsequent stage, the one with dinosaur skeleton and the old airplane, where instead he won and didn't skip the winning pose animation.

You can try to replicate those situations and make some comparison.

[model123]

Hello Harry

I tried to take a screenshot similar to the scanned image
But I did not get the same pose

Virtua fighter can choose a pose by pushing a button
But I could not choose it with Fighting Vipers

[Ian]

Hey Harry,
sorry it's taken me so long. Just looking at your last specular patch. Few questions.

If i was to guess what the specular values might be, they would probably look something like what you had before

Code: Select all

expIndex = vec4(8.0, 16.0, 32.0, 64.0);

How did you work out these?

Code: Select all

expIndex = vec4(9.0, 17.0, 40.0, 88.0);

Do you have some visual guides to compare the two against?

This is also very interesting

Code: Select all

         // (Experimental) 256 level posterization, mostly evident in Harley
         NdotL = round(NdotL * 255.0) / 255.0;

Should we maybe round to 128 if the math is being done with signed ints?

[HarryTuttle]

How did you work out these?

Good question

Well, I did by trial and error and visual observation and confirmation. The former values, although being more logical, were wrong. Basically I just "tweaked" the old ones to match the arcade. I've to still understand why on flat polygons those number don't work quite well.

That's why I think specular terms could be just hardcoded on Real3D, and shininess (0-3) is just an index to those.

Do you have some visual guides to compare the two against?

I'll post as soon I've some spare time, this will be a busy week for me, however the first games that come in mind now are LeMans24 and Daytona, plus some very subtle differences in Scud.

Should we maybe round to 128 if the math is being done with signed ints?

I thought of that too, I remember it became too exaggerated and "distracting" compared to the videos I was watching as reference. However this is the Harley example I was referring in my precedent post:


That screenshots is very old, by the way. Model3 is on the left, I was simulating that effect that doesn't seem a codec artifact at all. Also the horizontal "seam" on the glass I think is a quad poly interpolation artifact and not the effect of the capture card out-of-vertical-sync (which instead should "split" the whole frame).

Something similar seems present also in some Abelardo's videos of Daytona2 (the rear windshield of the player's car) especially in the city course.

[HarryTuttle]

This is how the red car looks like in my private build (256 posterization levels):

h4.jpeg (226.72 KiB) Viewed 8742 times

[Ian]

I'll try this out with my quad emulation code
You are right there is definitely slightly more banding in the model3 version (don't think that's a video artifact). But it's hard to say exactly where this banding is coming from with regards to losing precision.

[YoYo09]

Ian
R678 Virtual On
Background bug.


Oh, and the R678 has the Specular latest patch applied?

[Ian]

The sky is supposed to have stripes in it? That's what the original h/w looks like, unless I broke the fog code with my last commit.
I haven't merged the updated specular values yet, on the to do list.

[YoYo09]

The sky is supposed to have stripes in it? That's what the original h/w looks like, unless I broke the fog code with my last commit.
I haven't merged the updated specular values yet, on the to do list.

All previous screen shots are R678

This is a R677 screenshot



The R677 background is the same as the XBOX360 version.
(Exclude stripes)

So I thought it was a bug.

[Ian]

It might be a bug, make a save state and i'll try it out