Julien-Robert

I'm looking for a shader like td.lumadisplace.jxs with the function of taking the luma map of the first image to displace the second image.

Does anyone have heard about something like that?

luma-displacement-map-with-2-images

perhaps you could use td.repos.jxs to achieve similar effect, or you 

could try modifying lumadisplace shader to use two texture inputs.

perhaps you could use td.repos.jxs to achieve similar effect, or you  
could try modifying lumadisplace shader to use two texture inputs.

Thanks. The problem is that I don't understand what does the td.repos.jxs do exactly. The only doc I've found was inside the code: "distortion based on position map in second inlet"

Do you know what this shader does and how to use it?

Unfortunately, I'm not very good to program GLSL so it will be quite long for me to change the code of the lumadisplace for 2 textures...

Ok, I've found how to use the repos.jxs shader. That's almost what I was looking for isn't?

Yes that's exactly what you want to use:)

If you look at the td.repos.jxs in text editor, you'll see this line:

    vec4 look = texture2DRect(tex1,texcoord1);//sample repos texture

which assigns color&alpha value from second texture going into 

jit.gl.slab into a variable named 'look'.

texcoord1 is variable containing the coordinates of the pixel you want 

the value of 'look' is later used to modify texture coordinates of the 

    vec2 abso = look.ra*texdim0*amt;//absolute coordinates

which means 'repos coordinate is value of 'look' multiplied by the 

size of first texture(texdim0) and then multiplied by the amount 

look.ra means that shader uses only red&alpha channels.

this is where one of the differences between lumadisplace and repos 

takes place - repos uses red channel to modify X coordinate and alpha 

channel for Y coordinate - just like jit.repos.

    vec4 repos = texture2DRect(tex0, coord);

this is where the first texture(first inlet of jit.gl.slab) is being 

sampled for the final output, using the coordinates modified by the 

So, this is a ruff overview of what's going on in td.repos.jxs. If you 

want to modify it to use the luma value to offset both x&y 

coordinates, you would need to first calculate the luminance of the 

pixel, and then use it to offset both coordinates by the same amount.

Now you could take a look into saturation shader(cc.saturate.ip.jxs), 

    vec3 intensity = vec3 (dot(texColor, LumCoeff));

here they calculate the dot product between color and 

'grayish'(LumCoeff) values. Lumcoeef is defined earlier in the shader 

    const vec3 LumCoeff = vec3 (0.2125, 0.7154, 0.0721);

So you could use these two lines to create a modified version of 

td.repos.jxs - first you should duplicate the file and rename it to 

Then you can open td.lumarepos.jxs, and start hacking by pasting the 

But you should take care not to brake the structure of jitter shader, 

so first you could paste 'const' line after the 'uniform' declarations 

Next step is to add the dot product calculation to your new shader.

In repos shader, 'look' variable contains color information of the 

And this is what you can use to calculate the luminance(instead of 

texColor used in saturation shader), by adding this line right after 

    look = vec4 (dot(look, LumCoeff));

and now you have a lumadisplace&repos crossbred mutant shader!

But if you're not interested in learning glsl, then you can just 

insert 'cc.saturate.ip.jxs' slab with '@param alpha 0.' between your 

displacement texture and right inlet of slab contating td.repos.jxs 

On Jan 25, 2009, at 8:18 PM, Julien-Robert wrote:

> Ok, I've found how to use the repos.jxs shader. That's almost what I 

	vec4 look = texture2DRect(tex1,texcoord1);//sample repos texture

which assigns color&alpha value from second texture going into  
jit.gl.slab into a variable named 'look'.
texcoord1 is variable containing the coordinates of the pixel you want  
to read.

the value of 'look' is later used to modify texture coordinates of the  
first texture:

	vec2 abso = look.ra*texdim0*amt;//absolute coordinates

which means 'repos coordinate is value of 'look' multiplied by the  
size of first texture(texdim0) and then multiplied by the amount  
parameter.

look.ra means that shader uses only red&alpha channels.
this is where one of the differences between lumadisplace and repos  
takes place - repos uses red channel to modify X coordinate and alpha  
channel for Y coordinate - just like jit.repos.

	vec4 repos = texture2DRect(tex0, coord);

this is where the first texture(first inlet of jit.gl.slab) is being  
sampled for the final output, using the coordinates modified by the  
second texture.

So, this is a ruff overview of what's going on in td.repos.jxs. If you  
want to modify it to use the luma value to offset both x&y  
coordinates, you would need to first calculate the luminance of the  
pixel, and then use it to offset both coordinates by the same amount.

Now you could take a look into saturation shader(cc.saturate.ip.jxs),  
and see how luminance is calculated:

	vec3 intensity = vec3 (dot(texColor, LumCoeff));

here they calculate the dot product between color and  
'grayish'(LumCoeff) values. Lumcoeef is defined earlier in the shader  
as a constant value:

	const vec3 LumCoeff = vec3 (0.2125, 0.7154, 0.0721);

So you could use these two lines to create a modified version of  
td.repos.jxs - first you should duplicate the file and rename it to  
td.lumarepos.jxs.

Then you can open td.lumarepos.jxs, and start hacking by pasting the  
lines for calculating luminance.
But you should take care not to brake the structure of jitter shader,  
so first you could paste 'const' line after the 'uniform' declarations  
and before void main() function.

Next step is to add the dot product calculation to your new shader.
In repos shader, 'look' variable contains color information of the  
second texture:

And this is what you can use to calculate the luminance(instead of  
texColor used in saturation shader), by adding this line right after  
reading the texture:

	vec4 look = texture2DRect(tex1,texcoord1);//sample repos texture
	look = vec4 (dot(look, LumCoeff));

But if you're not interested in learning glsl, then you can just  
insert 'cc.saturate.ip.jxs' slab with '@param alpha 0.' between your  
displacement texture and right inlet of slab contating  td.repos.jxs  
to achieve the same result:)

>
> Ok, I've found how to use the repos.jxs shader. That's almost what I  
> was looking for isn't?
> --
> Julien-Robert
> jitter@cycling74.com`


Thanks a lot for all these explanations!


luma displacement map with 2 images