deleted comment, fake video was posted.

@evero 99% sure Blackmagic can take the back end of the 4K camera, stuff it in a box and charge $1000-$1500 for it, so we may see something come out that is less than the YAGH-a-meister.

The reason I was asking about downscaling before keying, is that I read somewhere that when you scale down, the color sampling is for some reason not scaled down as it is already downscaled in 4:2:0 - I don't know if that is true or not, but several people reported better keys when scaling down 1080p to 720p in 4:2:2 on the GH13. My tests with the GH13 was inconclusive, both using the 4:2:0 and 4:2:2 modes in te camera itself.

But @driftwood what I would love is just to see a keying example with the footage from the camera out of the "box" (not the YAGH...) Just a simple shot infront of a green screen and 1 click keylight example to see if it is any better than the GH's. Preferably in both good, adequate, bad and horrible conditions :) But anyting would suffice.

An extreme low-light example and also a challenging high contrast situation would also be cool to see how it handles.

@jamesgh2 That vid was a raw test shot on the pocket cinema camera, posted last year.

@driftwood - sorry, forgot to mention FPN - keep getting bit by my GH2 and some reassurance that GH4 is half decent in low light would be fantastic. Am assuming there's no iso bug either..

@Mistas & @AdamT

"I would stabilize the 4K footage so you don't lose resolution when the footage is cropped. Then downsample the cropped footage to 1080p;"

Stabilizing is extremely processor intensive. To add insult to injury, it won't scale with multi-core processors.

I would approach stabilization it in a different way. I'd first downscale the footage to something like 480p (4.5x smaller than 4K) and apply stabilizing to that. This should be ~20x faster to stabilize than 4K, but never mind the resolution, because you won't be using this footage. After you do this, you'll end up with a cropped/resizing factor to completely remove the borders at 480p.

Let's imagine that Mercalli/Premiere(WarpStabilizer) ends up with a 110% scaling (which is a bit above typical shake) to remove the black borders. Now you can easily compute the final rescaling factor to get stabilized FullHD from 4K = 50%(=4K/1080p)*110%(=crop factor) = 55%.

So this means that you can downscale first to 1188p, stabilize, and still get 1080p when the stabilizer crops 10% out to 1080p to remove the borders, without the need to do any upscaling nor processing unneeded data.

In the end, it will typically be 3x faster and will preserve the same exact quality as if you stabilized 4K and cropped after. ;)

Aerial Shot :

Whoa.....I have to say....I liked those whips and pans.

@jakepowell

Ok I removed the post thanks!

@duartix

The real essence of my question is will the quality of stabilization be better at 4k... Does more pixels = more info for software to analyze??? I realize it would take much longer and may not even be practical, it was more out of curiosity.

@amateur I'm not seeing any comparison between the two cameras in this video... I had to turn it off because of the annoying soundtrack... but this could have been shot with the GM1: not that that's a bad thing...

Edit: I went back and played the 'end credits': the entire video was shot with a GH3!

@driftwood

Can you do GH2 moon7 vs GH4 200mbps 1080p video? That would be interesting as a benchmark.