karl krach

I'm trying to do some sort of slitscan but with a moving video as a result, not an image... I'll try to explain:

I have a camera feed, eg. 1000 pixels wide, running at 25 fps. This I capture to a circular GPU buffer (with jit.gl.textureset.js), recording the last 20 seconds (= 500 frames).

What I want to do now, is split/concat these textures in a way, that my 1000 pixel wide image is composited out of 500 x 2-pixel-wide, vertical stripes, each pulled from a different frame in my video buffer - like a slitscan, but constantly updating at 25 fps.

How would I do this? I did search the forums quite a bit but submatrix does not seem to work with moving backgrounds and pretty all I found deals with 'frozen slitscans', not moving ones... what shaders/commands should I be looking at?

Hi,
I'm trying to do some sort of slitscan but with a moving video as a result, not an image... I'll try to explain:

I have a camera feed, eg. 1000 pixels wide, running at 25 fps. This I capture to a circular GPU buffer (with jit.gl.textureset.js), recording the last 20 seconds (= 500 frames).
What I want to do now, is split/concat these textures in a way, that my 1000 pixel wide image is composited out of 500 x 2-pixel-wide, vertical stripes, each pulled from a different frame in my video buffer - like a slitscan, but constantly updating at 25 fps.

As always, thanks a lot for any help!!!
k

Hi,
I'm trying to do some sort of slitscan but with a moving video as a result, not an image... I'll try to explain:
I have a camera feed, e

video-not-photo-slitscan

as you know i am not a jitter pro, but why dont you do it with a simple b&w masking?

I tried to wrap my head around this and cobbled some code together but I fail at imagining how i could possibly scale this...:

In the patch below I combine two textures in the same jit.window, each one I can freely mask/crop and position in the main output. 

The left one shows the current video input, the right one the output of a GPU frame buffer, so I can delay the input video. So far so good.

But, how the heck I would automate and scale this? As outlined above, my final aim is some sort of slitscan with like 500 or 1000 of these vertical slices all cobbled together... To get there I would need to

- find a way to generate and composite these different texture slices dynamically and

- read out my jit.gl.textureset.js not one frame at a time but like all of it every beat.

Or is my whole approach stupid and there is better ways to achieve this? I'm surely no openGL genius...

Hello again,
first @roman: thank you for your input!

I tried to wrap my head around this and cobbled some code together but I fail at imagining how i could possibly scale this...:
In the patch below I combine two textures in the same jit.window, each one I can freely mask/crop and position in the main output. 
The left one shows the current video input, the right one the output of a GPU frame buffer, so I can delay the input video. So far so good.

But, how the heck I would automate and scale this? As outlined above, my final aim is some sort of slitscan with like 500 or 1000 of these vertical slices all cobbled together... To get there I would need to
- find a way to generate and composite these different texture slices dynamically and
- read out my jit.gl.textureset.js not one frame at a time but like all of it every beat.

Or is my whole approach stupid and there is better ways to achieve this? I'm surely no openGL expert...

Hello again,
first @roman: thank you for your input!

I tried to wrap my head around this and cobbled some code together but I fail at imagining how i could possibly scale this...:


In the patch below I combine two textures in the same jit.window, each one I can freely mask/crop and position in the main output. 
The left one shows the current video input, the right one the output of a GPU frame buffer, so I can delay the input video. So far so good.

But, how the heck I would automate and scale this? As outlined above, my final aim is some sort of slitscan with like 500 or 1000 of these vertical slices all cobbled together... To get there I would need to
- find a way to generate and composite these different texture slices dynamically and
- read out my jit.gl.textureset.js not one frame at a time but like all of it every beat.

Or is my whole approach stupid and there is better ways to achieve this? I'm surely no openGL expert...

Thanks a lot for any help!!!
k


ok i suppose you dont only want to have a linear movement of a 1 pixel broad slit across the moving images of the other channel, to recreate how slitscan photography work, but rather make massively use of displacement, use interpolation between frames and all that. :)

depending on the material it might be less interesting than playing around with only one frame, and if you do it with moving images i am afraid you would need to create one stream per input frame, isnt it? for a given window, that is. say like 2 seconds. so that you are able to mix the modulated frames against each other.

ok no more wild guessing, need to look at the patch later.

again, thank you so much for helping me with this! i'm not sure i'll need interpolation and stuff but it certainly is a bit more complicated then the traditional 'photo slitscan'...

anyway: below you'll find my current patch, where i try to draw ten slices with a single gridshape which gets assigned differrent positions and textures. currently this does NOT work though as it seems only the last of these gridshapes is actually drawn to the render context...

also, i'm still not sure about the general approach - i feel there should be easier / more elegant ways of achieving this... if in doubt rather look at my old patch though which explains my basic problem better.

hi roman,
again, thank you so much for helping me with this! i'm not sure i'll need interpolation and stuff but it certainly is a bit more complicated then the traditional 'photo slitscan'...

anyway: below you'll find my current patch, where i try to draw ten slices with a single gridshape which gets assigned differrent positions and textures. currently this does NOT work though as it seems only the last of these gridshapes is actually drawn to the render context...

also, i'm still not sure about the general approach - i feel there should be easier / more elegant ways of achieving this... if in doubt rather look at my old patch though which explains my basic problem better.

again, thanks for any help!
k


did i explain my problem clear enough? did i miss something?

so, where are the gpu-gurus hiding?
did i explain my problem clear enough? did i miss something?
please, help!
k

if i understand correctly, you probably want a unique gridshape object per slice. you can use poly~ or JS or jit.gl.multiple, or simply make several gridshape objects or abstractions. use automatic or enable or simply an all black texture to hide them.

Video (not photo) Slitscan?