I’m having a problem making a garbage matte for a greenscreen sequence.
I have the greenscreen sequence and the rgbscreen1.8 plugin to key out the green.
Then I add a scene containing only a garbage matte and set it as alpha-over the green screen sequence.
So I have: scene - greenscreen sequence - alpha-over - rgbscreen plugin; and the matte won’t show. If I delete the plugin the matte appears.
It seems I can have either plugin or garbage matte, but not both.
If I use an image as a matte and select first the plugin, then the image matte and add alpha over, the matte renders, but as black, while the greenscreen sequence has transparency, and I can’t layer in the background sequence :-
I’m guessing maybe it just doesn’t work the same way anymore. Can anyone explain how it does work now?
Or, if I render the sequence with the image matte, how can I convert the rendered matte to transparency?
you need to make the material for the garbage matte plane R=0 G=0.8 B=0 and shadeless… you are actually superimposing a easily keyed matte over your greenscreen footage before you key it with the plugin
shut off OSA, Ray, SS, & shadow ( not needed and they just slow your scene renders down… in fact OSA can actually screw up the edges of your garbage matte)
Organic, out of interest did you start with a .tga image sequence or convert from a video format? Looks like the free greenscreen videos for download and then you converted it to tga with blender?
Hi paprmh, and thanks. I was hoping it would be something simple like that. I was in too much of a hurry, I guess I read it badly. I accept that was a pretty stupid error though. I better go read it a couple more times.
Possibly I am still confused, but I still can’t get the matte to show in the preview or to do its work in the render.
I can’t pretend I know anything about how the VSE works, but it seems I add the plugin to the footage, then add the scene, as an alph-over for the footage, and the plugin ignores the scene. There must be something else I have missed. :o http://www.mediafire.com/?28988wxodc9z1c8
yellow - Yes, it is the Hollywood Camera Work footage. I converted from a .png sequence. The original has a 1.5:1 pixel ratio, so I changed it before tracking it in Voodoo. Trying to figure out the film back with the ‘wrong’ pixel ratio was just too confusing for a noob. :o
I remember having a hard time getting some of the VSE plugins to work for 2.4x. Some were handy indeed but only worked with specific versions and were very crash-happy!
I remember the one I liked most (called old footage or sth) would crash blender instantly.
Btw, does anyone know if there are plans to port some/any/all of those scrips to 2.5?
I think that node presets would solve the problem.
@organic, I know that the plates are free but they’re pretty lowsy plates looking at the jagged edges to the talent. Not wishing to add any confusion to your query but you could also use the YCbCr or RGB separate node and a feed each channel through a blur node using cubic interpolation with values of 1 or 2 for x & y.
Then use a YCbCr or RGB combine node and feed the output it into your greenscreen plugin, it helps a bit with tidying the edges but really the source material you are using has been badly converted from movie format to .png in the first place.
Better would be to get some unaltered movie footage from another free place and convert that for better quality. Blurring it after a bad conversion is second best to decent source material.
I agree, the quality is rather poor, but I honestly couldn’t find anything even close to that resolution that was free.
From what I have been reading, that would be good approach, and I wish I understood it enough to just jump in and try it. I admit I am still pretty ignorant about the theory, though I might give it a try both ways, as it will be educational.
I just wish learning didn’t have to hurt.
Thanks blendercomp, that looks good. I will try to investigate a nodes method too. Do I need any special material settings for that plane?
Check this out, give them an email and you’ll get an auto reply for downloading 1280x720 RGB quicktime shots, I’ve looked and it’s far better quality than what you have.
I see the problem now, you are connecting the RGBscreen plugin to channel 3 ( the green clip ) instead of channel 4 ( the alpha over strip) Just select the alpha over, shft select the plugin and hit r
[ATTACH]115576[/ATTACH]
@yellow: the dv filter button on the plugin does a small blur on the U&V channels and does a very nice job on the jaggies…
@blendercomp: I did fix alot of the original plugins but nobody seemed to care… and as of now, there are no plans to bring plugins into 2.5
I hate to be a nuisance, but I tried it that way too. It looked good in the preview, and even in the render window, but once saved it looks like this:-
that screenshot looks like it is working… Did you actually hit F12 to render it before you saved? The preview window (the top right one) shows that everything is working correctly, If you hit F12 then F3 you should get what you see there…
I also tried the Anim button and autosave, but with the same result. I can’t figure it out. Of course it doesn’t help that this is new to me, I’m doing this to learn how.
ah… it’s not you, it’s a bug. I never noticed it before but I don’t use scenes in the VSE - it slows things down too much. Apparently, if you have a scene strip in the VSE, it will only save the scene render…weird… I know that it worked before.
Anyway, all you have to do is get rid of the scene strip and just use the rgba .png of the mask instead.
That’s useful to know, haven’t tried it out in a long time. With regard to going the nodes route which was the context, not sure if a blur option is available, I’ll check it out.
Out of interest does your plugin for the VSE take the frame straight from FFMPEG in YCbCr or from Blender?
@yellow: the footage is incredibly crappy. This is the main reason why they have released it. I’ve played around with it and never really managed to pull a decent matte. Most of the compositions I have tried sucked big time. At first I thought it was just me, but then I read an article on their site in which they explained all the shooting errors. The only reasonable quality composition I’ve seen using that footage was in a ramen showcase.
@paprmh: thanx for keeping the scripts up to date; I guess that if they are ported to 2.5 some users might find them useful. I know for a fact that users would prefer VSE addons rather than having to learn compositing and the node editor.
If I understand what you’re saying, in the compositor one would blur the matte before continuing with the noodle. That would usually smoothen edges and create a nicer composition.
Part of the poor quality is the conversion to .png’s. But it’s a challenge I guess.
If I understand what you’re saying, in the compositor one would blur the matte before continuing with the noodle. That would usually smoothen edges and create a nicer composition.
Well not quite, before pulling the matte, it was specific to the jaggies really. The plugin does that however, re paprmh comment.
Blurring the edge to the matte would follow later.
