Best method for compositing a CG scene with actors

[DryMeat]

Hi,

I've been wondering what the best or at least ideal way to go about compositing a cg scene with live actors. I am still fairly new when it comes to compositing for video in general. I have been going through some of the Maya and after effects lessons here with the road work sign. My question is what other courses here can anyone recommend that will help me quickly learn to accomplish this. I'm working with others on a indie short film sort of sci fi ish which will be mostly if not entirely done on a green screen backdrop. Basically i need to model a city street/intersection where a fight takes place. I understand that nuke is a better suited compositor for heavy cg compositing. Correct me if i am wrong by the way. I believe that adobe premier pro and after effects will be mostly used for this project so i don't know if trying to learn how to use nuke for now would be wise in dealing with deadlines. Also will Dt in the future offer lessons on video editing with premier pro and/or other video editing programs ? Just curious.

Thanks for any help

[KMKaine]

Hi Drymeat,

I Think After Effects will do just fine for what you intend. As it sounds to me it is quite 'basic' compositing work, and where Nuke indeed has great possibilities, it probably will be overkill. And as you state, I think going with After Effects initially is a wise decision. Also for overal compatibilty (more people have After Effects than Nuke).

Something you have to put a lot of time and effort in is the green/bluescreens. Make sure you have good lit, well thought over greenscreen setups. Having good greenscreens makes your life a whole lot easier! Evenly lit, good green, try to reduce reflections from the actors clothing (like belt buckles and jewelry), and the format you use for capture can help you a great deal...

Shooting on DV isn't ideal for example when using greenscreens. Nothing is impossible, but using something with a little less color compression is better (Also the D7 or D5 aren't ideal for that) try using Digital Betacam (HD) or Red for example, but I can imagine that there isn't a lot of budget, so ofcourse you have to make do with what you got. When using DV or HDV or something similar it is even more essential to setup up good greenscreens. As your entire movie is going to be shot in front of greenscreens, it essential that you work that out well. It could be wise to shoot a test scene / shots so you know if the setup wil work. That way you can eliminate problems when you run into them while working on the testshots, and apply the fix when shooting the real deal.

Also don't forget to put up enough trackermarkers (if you are going to use moving shots, and matchmoving) and I'd suggest doing some tests with that as well.

And a simple but often made mistake: Pulling a key, doesn't have to work with one click ;-) Most of the time you get better keys when combining multiple 'key passes' ;-) For example, pulling some garbage mattes, then hard key to get rid of the green, and then make a new key using some softer settings so that the the actor has nice soft edges and blends better with the background (ofcourse not to soft, because then it's obvious the actor is comped in ;-) )

Hope this helps a bit, and let us know if there is anything else....

[bravo2zero]

The software is the lest of your problems you could use paintshop to overlay frames and renderings to get what you want . Whats more important is the steps you take to achieve your finished work .

I can remember back in the day cutting out masks then applying the cg to each frame one at a time .

How ever maya allows us to project images onto a background for ref so if your working on 1 mins of video , we can set our time slider to 1920 frames roughly 1 mins at 32fps .
then take the 1 mins of real footage we want to work with and brake it down into frames and project each frame onto the background and bind each one to the time slider so our background changes slightly from one frame to the next so we can get objects positions correct . this would be a good job for mel else it will be repetitive .
How ever getting the depth of field correct can take a bit to get right but that's part of vfx a lot of trial and error . i suggest you dont do more then 32 frames at a time or 1 second else it will seem like a massive endless task .

All the above can be done in just 3-4 lines of mel and using ae to brake down the 1 min of footage into frames . this then leaves you plenty of time to get stuck into the rigging of objects .

Then by turning of your reference off (eg background) layer take your 1920 renders frames with mask and overlay them on the original frames in ae and your very close to being done with out paint lighting etc.

[DryMeat]

Wow, it seems like this is much for involved than i figured. I'm glad i wont be the only one working on the cg/vfx portion of this. Hopefully when other artist come on board i will be able to pick up things and learn how they go about these tasks as well. There was talk about using a test shot to get an idea of how the initial run will go. I expect more details on the studio setup and crew capabilities as it gets closer to begin shooting. Thank for the info guys.

[KMKaine]

Hi Bravo,

I don't really understand what you mean exactly? Are you talking about camera projection? Or rotoscoping? I think DryMeat wil probably be justing Matchmoving to match real camera with the CG, and then Key in the actors?! When using the greenscreen there isn't much need for rotoscoping I guess...?

Let us know how it works out DryMeat!

[bravo2zero]

When you shoot footage with actors you might get 3 takes the best take is brought into the vfx department and from this footage master slides are created. and that's all the footage we have to work from . when adding in cg elements we cant re shoot the footage as there is a substantial cost involved in bringing actors back to a set plus setting up a set that generally gets torn down after the shoot .

So from our master slides we have to make the cg work with what we have not the other way around . Call it what you like , But we use the master slides as ref when animating to make our cg witch is cheap to manipulate , work with the master slides. then put the animation as an overlay over our master slides or to add something further behind and add the master over the cg layer we dont have much green screen footage to work with when most is shot on location . most green screen shots tend to be when its to dangerous for people to shoot it for real , situations where people are falling grate distances or having trucks and cars explode near people of bringing down buildings etc most of these kind of shots are built up of many layers, with a background then a layer with the a person and then the fly debris.