Beta release 2.0.0b

[Kjell]

Another important animation feature .. morph. Since the current Mesh component doesn't expose the vertex identity, all rendering is done using opengl32. Attached is a basic example using linear interpolation and 2 targets ( + base ) .. normally you'd use this in combination with F-Curves for facial animation / muscular deformation etc.

K

[Kjell]

Just messing around .. bloom shader based on gaussian blur & FBO mipmap ( + effect in overdrive ).

K

[jph_wacheski]

Hay, that looks real nice! Care to share the .zgeproj for the curious?

[VilleK]

Yeah Kjell you are just teasing us with your skills if you won't share the source
Or at least an exe so we can see it in motion. The stills look very nice!

[Kjell]

Reason I didn't post a exe was because I have no idea whether it works on a ATI card or not ( and don't need to know either ), plus the effect is not resolution independent ( and there's a issue with the AspectRatio of FBO's using AutoPowerOfTwo ) .. just experimenting with different techniques.

Anyway, see attached.

K

[jph_wacheski]

Does not work on ATI,. at least here.
Stupid vidcard inconsistancies.

[Kjell]

My guess is that this is similar to the Accumulation / Stencil situation. Since ZGE doesn't explicitly enables these features ( FBO mipmaps in this case ), they will fail on ATI .. while NVIDIA is a little more forgiving.

K

[VilleK]

It does not work here either. It displays a small rectangle in the middle of the screen where you can barely see moving colored cubes that looks like those in the picture.

What is the FBO mipmap issue? Zge already makes the glGenerateMipmapEXT-call which is supposedly only needed on ATI.

[Kjell]

Are you sure? Google doesn't find anything when I search for glGenerateMipmapEXT in the repository.

K

[VilleK]

That's odd because searching is what google usually do best

If you do a normal google search for "glGenerateMipmapEXT zgameeditor" it finds the code:

http://code.google.com/p/zgameeditor/so ... nderer.pas

[Kjell]

Ok,

Then I might have scr*wed up somewhere .. here's the fragment shader ( perhaps u / v / offset need to be part of main .. or bloom set to vec4(0.0) ).

Code: Select all

uniform sampler2D tex1;

uniform float canvasWidth;
uniform float canvasHeight;

const float kernel[9] = {1.0/12.0,1.5/12.0,1.0/12.0,
                         1.5/12.0,2.0/12.0,1.5/12.0,
                         1.0/12.0,1.5/12.0,1.0/12.0};

float u = 16.0/canvasWidth;
float v = 16.0/canvasHeight;

vec2 offset[9] = {vec2(-u,  v),vec2(0.0,  v),vec2(u,  v),
                  vec2(-u,0.0),vec2(0.0,0.0),vec2(u,0.0),
                  vec2(-u, -v),vec2(0.0, -v),vec2(u, -v)};

void main()
{
  vec4 bloom;

  for(int i=0; i<9; ++i)
  {
    bloom += texture2D(tex1, gl_TexCoord[0].xy+offset[i],4.0)*kernel[i];
  }

  gl_FragColor = texture2D(tex1, gl_TexCoord[0].xy)+bloom*4.0;
}

K

[VilleK]

It complains about the syntax of initializing constant arrays.

This compiles ok:

Code: Select all

//...

const float kernel[9] = float[](1.0/12.0,1.5/12.0,1.0/12.0,
                         1.5/12.0,2.0/12.0,1.5/12.0,
                         1.0/12.0,1.5/12.0,1.0/12.0);

//...

vec2 offset[9] = vec2[](vec2(-u,  v),vec2(0.0,  v),vec2(u,  v),
                  vec2(-u,0.0),vec2(0.0,0.0),vec2(u,0.0),
                  vec2(-u, -v),vec2(0.0, -v),vec2(u, -v));

[Kjell]

Ok,

Guess that's good to know

Something completely different, quick experiment using a QTVR type of approach. Move with left mouse button, look around with right.

K

[VilleK]

Haha, that's a very cool effect. Works fine here.

[y offs et]

Watch your step!
Works here too, although minor jerky.
7Z seems to take an unusual long time to unzip it.