Tricky compositing infrared shots

[Evan Thomas phillips]

I don’t know how much keying you’ve done from things shot on MiniDV, but it isn’t easy, especially if you’re ingesting the footage via firewire into your NLE. This will cause aliasing around your subject that makes keying almost impossible (unless you’re going for a certain look that will allow a massive alpha halo). What I’ve done in the past, is use a BlackMagic Multibridge to import the film via Analog Component as Uncompressed 10bit 4:2:2. Your files will be a lot bigger, but it smooths out the color channels and gives your footage (and keys) a more natural look. After that, I usually use KeyLight in AfterEffects; it does a pretty good job without a lot of fuss. The Multibridge is about $1500, or you can pay a post production house $200/hr. Good luck.

[Jonathon Sendall]

OK, then the option would be to shoot in something like HDV, do the keying etc and then dirty it down to look mini-dv. Seems a bit perverse to go that way but if it makes the post simpler then that’s the way to go.

Story, not pixels.

Jonathon Sendall

[Bbalser]

You could just shoot in the pool and use Shake to mask out the body. Not too difficult, but you do need to know Shake.

[Evan Thomas phillips]

Actually, HDV is even more difficult to key than regular DV. The problem is the amount of color information, not resolution. Professional video equipment captures at 4:2:2 color space, which is pretty easy to key if it’s lit right. DV is only 4:1:1, it still has some Chroma information to draw from, but only half of what you really want. HDV is a 4:2:0 codec, which basically trashes the Chroma information; that’s how the HDV codec achieves 1080 lines of resolution at only 3mb/s. The reason to go analog into your NLE (you might even be able to rent an AJA IO from someplace) is to upconvert the DV 4:1:1 to 4:2:2 color space. It basically fakes the chroma information.

Another option would be to key on Luma, instead of chroma, but that would take some very precise lighting to pull it off, and you would probably want Shake to do the compositing.

[Jonathon Sendall]

So one more question. What if I shoot DV, up to a better format, post the effect and then back to DV. IS there any benefit in that or will that be a waste of time? I know there are some converters out there that will anti alias mini DV quite nicely on the conversion. Anyhoo, thanks for the up.

JPS

Story, not pixels.

Jonathon Sendall

[Evan Thomas phillips]

If you up to a better format to do the effects, render it out as 4:2:2 Uncompressed 10-bit, then compress it back down to DV, you shouldn’t “lose” the effect. I know this process sounds crazy, and I would have never believed it myself unless I had actually done it. I suspect the issue arises because DV cameras capture an MPEG stream onto DV tape, which is a compressed signal, then to do a firewire capture it is compressed again with the DV/DVCPRO codec. Good luck.