laserpilot

I have a rather large video performance patch with a lot of effects (just posted as project 53 on the community projects page, thanks guys!), but I'm starting to get tired of turning on an effect and having it effect the entire frame.

Does anyone have a creative solution to applying filters to certain parts of the image? Of course there are simple things like thresholding, lumakey, chromakey etc etc but I have had terrible luck with alphachannel masking with my opengl effects..the jit.alphamask always works well, but not as much for the lumakey alphaglue shaders for some reason

anyway..I'm envisioning something like have an ipad/ipod touch interface where I can push a square of a blur or color filter or whatever around an image with 4 fingers which would be possible and I can think of ways to do it, but not sure if these would be the 'proper' way..also cropping in jitter seems to be lacking..there is a decent imageunit for it, but I don't know if I'm missing something where it is a common feature on jit.matrix that I just haven't explored yet

Mostly just rambling, but if anyone has some interesting ideas..I'm looking to expand my work past effects that just blast onto the entire frame and are a bit more subtle

cropping-filtersmanipulating-non-full-frame-filters

Sounds like what you are talking about is applying a travelling matte to effects in your system. You can do this in numerous ways, but the general idea is that you will need to composite a clean channel of video with the processed video, using some sort of alpha channel generation. If you have a patch that shows what you are trying to do, we can probably help you get it going.

Sounds like what you are talking about is applying a travelling matte to effects in your system.  You can do this in numerous ways, but the general idea is that you will need to composite a clean channel of video with the processed video, using some sort of alpha channel generation.  If you have a patch that shows what you are trying to do, we can probably help you get it going.

Finally got some time to mess with this more consistently. Here is the basic idea I am thinking of. If all goes right, when you turn on your camera, only the bright parts of what the camera sees will apply a blur on the dishes..you can try it with other effects in my slab chain as well.

What is the simplest way to composite these that doesn't involve spitting it out to a videoplane? Do i just have to capture them to a texture and do some kind of blend mode? I was trying to use things like co.additive.jxs to try and blend my masked video with the original but it wasn't working correctly. Things didn't look right until I turned on blend enable on the two videoplanes. Any other interesting ideas to try for masks besides camera input?

Nothing yet...I tried passing them into textures or meshes, but I think I'm missing some core piece of information about the proper way to pass things out of the two pieces to mix them properly..here is a slightly cleaner patch with both motion detection and luminance based mixing..still looking to add this ability to a slab chain and NOT have it end in 2 gl.videoplanes..

just send the final texture and the camera input matrix to a final compositing slab.

for a list of available compositing shaders, check out:

Max5/examples/jitter-examples/render/slab-helpers/composite.shaders-help.maxpat

otherwise, blending two videoplanes seems like a fine solution. what's the problem?

just send the final texture and the camera input matrix to a final compositing slab.
for a list of available compositing shaders, check out:
Max5/examples/jitter-examples/render/slab-helpers/composite.shaders-help.maxpat

otherwise, blending two videoplanes seems like a fine solution. what's the problem?


I was trying to send them to a compositing slab initially, but none of them seemed to cooperate in the same way as engaging blend enable on the videoplanes..they were always adding things together improperly or showing too much of the white background from alphaglue.

Is there some issue with how the composite shaders deal with alphachannels generated by alphaglue? I tried using some of the math shaders too but didn't get any decent results.I just want to add the original video to the video put out from alphaglue shader (which appareas to be a white background until things are revealed out of movement or luminance). I thought I tried every composite shader and didnt get the right result..but maybe I'll run through them all again..

The reason I didn't want to jump everything out to videoplanes yet is because these effects would be in a long line of the slab chain in my performance program, and at the end of that chain they still go to a gl.mesh for quad-warping for multiple planes..I have been assuming that once it goes to a videoplane, that is the end of the line for processing.

I just want to add the original video to the video put out from alphaglue shader

you can also try adding a scaling stage to each input before the final additive composite.

I guess "add" isn't quite what I meant but rather combine..I'm not looking for the blown out look from the additive blending. Attached is a version that switches (hit the big purple toggle) between what I want things to look like, and how they look when i use the additive blending slab at 0.5mix..it's just not quite right... If you aren't seeing what I am, I can post some screenshots as well

I'm not sure what you mean by scaling stage..like luminance scaling or something else?

Ah..well damn..it was co.alphablend.jxs ..thats what I get for relying on the shader helper which normally leaves that one out of the umenu list..I've never used it before..but this time..it worked! Wooo

If anyone wants the completed patch posted, let me know..or you can just replace co.additive.jxs with co.alphablend.jxs in my above patch..then drop stuff in your slab chain cleverly

Cropping filters/manipulating non-full frame filters