ARRI Burbank FBI Raid?

Erich Ocean said:
With RED, you can only do the debayer/LogC/ProRes conversion *after* the footage has already been wavelet compressed. The plasticky skin people hate with RED footage? I'm almost certain that's an artifact of the wavelet compression, not the sensor itself. With Alexa, the conversion happens in camera from the full, uncompressed ARRIRAW footage -- and is immediately written out as a super-high quality, 12-bit RGB LogC-encoded ProRes444 file.
Click to expand...

I think if you're getting plasticky skin tones, your lighting, exposure settings and/or processing of the footage is more to blame...
 
Stephen Gentle said:
I think if you're getting plasticky skin tones, your lighting, exposure settings and/or processing of the footage is more to blame...
Click to expand...

You seem to believe that lossy wavelet compression isn't a problem, or if it is, that it's not a problem with skin tones, or if it is, that it's a problem that can be overcome by "lighting, exposure settings, and/or processing of the footage."

Nonsense. Might I suggest that you learn the mathematics behind wavelet compression, examine the way skin is represented in a Bayer mask, and then get back to us?

Wavelet compression works by rounding away fine detail. Most of the time, that's not objectionable. But it smooths things out, and skin in particular is an EXTREMELY subtle and complex material with very fine detail in a very narrow range of values that does not interact well with wavelet compression. The result? Skin that looks plasticky.

And it's not just me that thinks so: http://www.cinematography.com/index.php?showtopic=46954&hl=alexa&st=0

This might help your thinking on the topic. You know how skin looks like crap in 3D rendering? That it, in fact, looks like plastic? Well, that's because skin is an extremely detailed and complex material with really complex interactions with light, in a very narrow range of values. The RED camera is getting good values on the sensor (IMO), but is then throwing them away in the wavelet compression. The result is like a 3D rendering -- the complex, fine details are lost. Obviously RED footage looks much better than a 3D rendering, but -- and this is key -- skin is not only really sensitive to minor changes in how it's rendered, but PEOPLE are really good at recognizing skin. For all I know, there's a special part of the brain that helps us identify skin. The end result is that *any* deviation from excellent skin reproduction get noticed quickly by people -- especially imaging professionals.

This aspect of RED footage is not controversial. Some people are okay with it (for the price), some people like it, but basically no one acts like RED doesn't have it's own look -- and that the look can be described as "plasticky".
 
Last edited:
Sorry Erich, but Stephen is correct. The way you light, expose, process and grade the footage has more of a bearing on getting plasticky skin tones, than whether the camera is shooting ProRres444 or REDCODE.

Look at the trailer for Knowing, shot on the Red:
http://trailers.apple.com/trailers/summit/knowing/hd/ – the skintones look monotone and plasticky right? Just like you said...

Compare it to the trailer to Leaves of Grass, also shot on the Red:
http://trailers.apple.com/trailers/independent/leavesofgrass/ – The skintones look a lot better.

The same camera, the same codec, two very different results.
 
Erich,

Last I checked, Cineform is a lossy wavelet codec, and I don't hear much complaining about the skin tones coming from it. Are you saying that you see the same loss in post with Cineform? Or in the SI-2K? Ir is this just a phenomenon of Redcode?
 
tonybuzzan said:
Sorry Erich, but Stephen is correct. The way you light, expose, process and grade the footage has more of a bearing on getting plasticky skin tones, than whether the camera is shooting ProRres444 or REDCODE.

Look at the trailer for Knowing, shot on the Red:
http://trailers.apple.com/trailers/summit/knowing/hd/ – the skintones look monotone and plasticky right? Just like you said...

Compare it to the trailer to Leaves of Grass, also shot on the Red:
http://trailers.apple.com/trailers/independent/leavesofgrass/ – The skintones look a lot better.

The same camera, the same codec, two very different results.
Click to expand...

"Knowing" also may have gone through a degree of digital noise reduction which often results in waxy looking skin. Too many variables to make a definitive statement.
 
PerroneFord said:
Last I checked, Cineform is a lossy wavelet codec, and I don't hear much complaining about the skin tones coming from it. Are you saying that you see the same loss in post with Cineform? Or in the SI-2K? Ir is this just a phenomenon of Redcode?
Click to expand...

I believe it is just a REDCODE phenomenon (*as currently implemented; Epic has a bunch of new compression options). There are multiple kinds of wavelet compression. Cineform "is based on an integer reversible wavelet compression kernel, with non-linear quantizer to achieve higher compression." (From Wikipedia)

Integer kernels are reversible, i.e. lossless. The compression ability in Cineform is coming from the quantizer, which IMO is a better way to remove redundancy (which is what enables the compression). It's sort of similar to how LogC is a better way to fill 12 bits than linear is, if that makes sense.

And to everyone else, um, no kidding that RED footage looks better when properly lit and graded. I'm surprised anyone here has taken the time to argue that. What's next? Maybe someone will point out that sound is better when you use something other than the on-camera mic?

We're talking about subtle problems rendering skin with RED footage shot *at it's best*. Anyway, it's all academic now; it can't be fixed on current RED One cameras, and when the Epic ships, we'll actually get to see how REDCODE is affecting the image because there will be a lossless option. My comments are directed towards people who have written off RED because the recorded image doesn't meet their standards -- and Alexa does. I'm trying to describe why Alexa is currently beating RED, and how the Epic will again even the score, or even surpass Alexa thanks to the 5K sensor.

I'm convinced it's not the MX sensor that's to blame. I hope I'm not wrong, because the sensor isn't going anywhere anytime soon, and it does have issues (although it's much better than the original RED One sensor). Even The Social Network, as good as it looks (and it's arguably as good as any RED One MX footage is ever going to look) still has that RED look to it, primarily in the skin tones. I think lossless Epic will truly push the MX sensor into the film replacement category on an aesthetic basis. Alexa seems to already be there.

Full disclosure: I've got a deposit down on a Stage 2 Epic already. I like RED a lot. But I'm also not oblivious to the areas where RED footage isn't quite matching what's possible with film. Yet. :-)
 
Last edited:
Warning: Pointed Comments Follow!!!!!!!!!!





Am I the only one that is shaking their head on much the TOPIC got diverted here? Uh FBI, Arri, Raid?

I could be a bit off in my synopsis, but pretty sure codecs, the Alexa, Cineform, REDCODE, and such had absolutely nothing to do with the raid, which is what I am interested in here.

Let's check: Thread title:
ARRI Burbank FBI Raid?
Click to expand...

No, I'm good.

My joke, was just that, a joke. Still want hard info. Hmmmm?
 
Last edited:
Supposedly ARRI is doing a press release tomorrow. They wrote it today and their lawyers are going over it with a fine-toothed comb.
 
You must log in or register to reply here.
Back
Top