Yet another light saber key test

I’ve made a quick test with some help from my youngest brother. Nodes all the way! In the future I need to make it so that my red actually looks like red to the camera because I keep ending up with these colors that are hard to key. http://www.youtube.com/watch?v=Ooxi0LlPLa0

Nice! :open_mouth:
How did you exactly do that? Give me link to the tutorial you used, please.

I didn’t use a tutorial. I didn’t save the blend ether. It was a 20 minute test. If you would like, I could try to duplicate the node configuration and post a picture.

Yes, please do!

This isn’t the exact set up I used but it’s pretty close and gets the job done.

Attachments


This still dosen’t explane how you were able to apply this to actual video.

I assume that a tracking program was used, such as Icarus, to motion track the movie and used the Icarus import script for Blender to apply it to the camera. From there, all that needs to be done is make the model and put it at the right place at the right time.

Tuts for Icarus can be found at: http://www.peerlessproductions.com/tuts.html

No blender never looked at anything in 3d so no motion tracking was used. The only difference was that the input node was a movie not a single frame. After that, I just hit “do composite” under the render buttons and it was rendered. Blender never looked at anything 3d, it didn’t even matter what the camera was looking at. You would have seen the default start cube if that was the case.

Brilliant amoose! Thanks for showing us, very useful. I take it that the lightsaber is a stick painted green?

I just HAD to try it, worked like a charm! Cool!

Gonna have to play with this a bit. :wink:

how did you mask out the background is what I wanna know…

Really Brilliant vid!
(edited to save face)

I suppose matching it up frame by frame would work. Personally, I prefer motion tracking it and applying it to the camera.

But that works too.

Cheers,
John

Frame by frame does work, but it is not fun! A year ago, some friends and I made a mini starwars film. I didn’t have Blender at the time so I used Macromediea Flash to apply the effect to each frame. It’s not fun at all! I spent over a month doing this tedious work and it didn’t turn out too well ether. So anything that can render it automaticly would be fantastic.

wow…that is so familiar, i can even explain it. :slight_smile:
http://blenderartists.org/forum/showthread.php?t=62611

One thing I want to clear up. You dont have to have green-ness for comping. Its generally used to tell teh computer what to take out of the video.

I take it that the lightsaber is a stick painted green?
That is very true. I didn’t use a stick painted green. I used one of those cheep Walmart light sabers that happens to be laying on a toy pirate ship. In this case it was a red screen. It came out kind of dark red, not the easiest color to key.

how did you mask out the background is what I wanna know…
I didn’t in my video. If you want to do this just put a another color screen behind the person and key it too.

Heres a simple way to do it(requiring nodes only for putting the render over the video). NOTE: This doesnt include intsruction on animation where the saber passes behind the Jedi fellow.

  1. you should see you(and your saber handle), and your 3D saber. Proceed to match up the 3D saber to the live-action one. Go frame by frame(or every ten, but then youll have to go back and clean it up) and match the two up and keyframe every frame till the end of your video.
    You have fun with that. I will must certainly not produce an effect that takes months to produce and is possible to automatically be produced in 30 minutes. I have tried the effect above last year. It doesn’t look very good on low quality video and takes a really long time to make on anything 1000 frames plus. Something that takes that much positioning could hardly be called simple.

Also, when the blade is moving fast it leaves a large blur of white surrounded by colored aura.
http://worcestermovies.com/wp-content/uploads/2007/05/rvd.jpg
This effect is hard to make using the method above. Motion blur doesn’t work. I’ve tried that.

On my method:
If you were keying a glow stick, or something that makes that effect in real life, the effect will be produced.

I suppose matching it up frame by frame would work. Personally, I prefer motion tracking it and applying it to the camera.

But that works too.
Do tell more. I don’t understand motion tracking can track a moving object as fast as a light saber. :confused: Generally motion tracking tracks the cameras movement not the subjects that are moving within the footage. Now if you are going to move the blade frame by frame, it does seam that your light saber movement would come out more correct to what is happening in the movie if you start out with a motion tracking camera.

wow…that is so familiar, i can even explain it. :slight_smile:
http://blenderartists.org/forum/showthread.php?t=62611
Right, it would have been quicker for free_ality to have posted the link.

Ha, thanks for the link Guys, I couldnt remember where I learned it :slight_smile:

And Im not saying your way is in any way bad, it looks absolutely fantastic. Just giving a nice way to do it basicly.

But yeah, you’re right, Motion tracking doesnt track objects, just camera moves…

Ha, overall, I leave this looking like na idiot :smiley: Such is life…

Drew

True but, if I could make it track objects or even locations on objects that would be nice. Like a python script that takes the mean location center of all the pixels that come though the screening process, and translates that into ipo curve…

That would be amazing, but would it be able to track the depth axis? They’re just abrely making cameras that can record that information…That’d be the toughest part…

Hmmm… maybe a second camera? If it was a light saber blade, and if you know the distance between the points and one of the points doesn’t move, you can triangulate since you know the length of two of the sides. Or maybe you can tell by the number pixels that make it through the screening process. (The closer the spot is to the camera the more pixels it will occupy.) Maybe this could be done with a wiimote? They spit out distance and location. I remember a thread on this in the past. I’ll look for it.