ML, Video Input and GLSL, texture2D vs. texture2DRect

gtoledo3's picture

I was looking into a "problem" that popped up with QC upon Mountain Lion, and just realized there is really no problem.

Some old compositions that feature a Video Input patch that directly outputs it's image to a GLSL shader seem to have texture coordinates screwed up now. I had been "working around" this by popping a Core Image patch in between, which sets the proper image attributes.

This was bothering me, so I did some testing and found a really simple answer:

Video Input now outputs a texture2DRect type texture, not texture2D. This of course changes the coordinate system from normalized to pixel w/h, thus explaining the look of something that isn't UV mapped.

Three here are 2 possible fixes, in levels of correctness (IMO):

1.-Make your texture in the shader a rect, not just texture2D.... so, ala:

uniform sampler2DRect texture;
void main()
   //Multiply color by texture
   gl_FragColor = gl_Color * texture2DRect(texture, gl_TexCoord[0].xy);


2.-Pass the texture through a Core Image kernel, which will also change the "2DRect" texture to "2D".

gtoledo3's picture
Re: ML, Video Input and GLSL, texture2D vs. texture2DRect

Can't edit the post for some reason... I should have phrased it that "there's not much of a problem". I don't want to take anything away from the fact that it's not that intuitive, and it used to "just work".