(Page 2 of 3 pages for this article  <  1 2 3 >)

Filed under:

5 Tips to Maintain Sanity in RED Post

Tip 2: Decide who, if anyone, controls the color intent of your source footage at the beginning

Many RED shoots treat the footage as if it were film; what is seen on the monitors at the shoot is the equivalent of a video tap that gives you a rough guide but not the look of the final image, like you might have with a standard 10 bit 3 CCD video camera. This is because the R3D file explicitly does not offer control of color intent; quite the contrary, essential information such as the ASA/ISO and source and target color space are essentially arbitrary (and out of your control if you pull the R3D directly into, say, an Adobe pipeline).This has a disadvantage, as RED does not have film’s dynamic range, so a light meter won’t necessarily pick up highlights that could clip.

Many DP’s know how to avoid clipping digital video, but will not be so ready to let go of shooting footage with the expectation of how it will look in post-production. It is certainly possible to preview RED camera output as REC 709 video and then make that the intended color space (in a tool such as REDAlert); you can even create a color look in REDAlert or Scratch and transfer it to an SD card to load it directly onto the camera as a preview LUT.

Many people don’t know this, and so it is the rare shoot on which this happens. Having now tested the use of SD-cards for this purpose hands-on, with a Build 16 RED camera proved that, as of now, you can preview Rec709, but no other output profiles, and color profiles cannot be transferred from REDAlert to the camera - or even from the camera to REDAlert - but only from one RED camera to another. RED has strongly hinted this will appear in a future build. Thanks to reader Steve Harnell for the tip.

 

If, in the more typical scenario, a shooter was going for a particular look on a REC 709 (HD) monitor, you can at least assign that profile is assigned to the footage. Here’s where it gets tricky - this is one subject that goes beyond the scope of this article: in REDAlert, you can specify a Color Space that corresponds to that of the shoot,  load or save a preset with specific settings, and those settings can ride alongside the R3D file as an RSX file; they are applied, for example, to the QuickTime files that point to that R3D file, so if you open the _P, _H or _M files that are also riding along, boom, you get a QuickTime with color adjustments as soon as you rewrite them (which you can also do with REDAlert) so that they are ready to use with color intent inside applications that work with those QuickTimes, such as Final Cut Pro. There are enough snags with this process that it is more appropriately saved for a more specific follow-up in a later article. Watch this space.

However, many other places that R3D file could go, including straight into After Effects or Premiere, do not recognize that RSX file and offer effectively only the raw R3D data plus a few rudimentary controls over grain and so on (covered in tips 3 and 4).

 

If I’ve lost you with this talk about REDAlert, hang on; I’ll try to clarify somewhere around tip 5.

 

Tip 3: Leave R3D and 4k behind as soon as possible (and no sooner)

This tip is the very heart of my argument here, and no doubt it is the one that will cause the most internal strife for conscientious video nerds.

Repeat after me:


I cannot output 4K R3D files. The longer I drag around huge entire R3D files when my actual plan is to scale them down (particularly to SD), the more misery I cause myself.

True story: a colleague for whom I have the utmost respect (name omitted to protect the innocent) posted a standard def, D1, 720 x 486 commercial full of medium close-ups perfectly shot on greenscreen with source 4K R3D files natively in After Effects. Despite that the images contained over 25 times more image data than his output required, he and his team evidently fell in love with how great the keys looked. When his renders failed he forced memory settings in After Effects to use only the disk cache for memory (because he was outstripping the 3 GB per processor RAM limit for this 32-bit app) and the renders then, instead of just failing, merely required several hours.

For 15 seconds of SD broadcast video. It didn’t take me that long to render a good greenscreen on a beige Mac in 1996.

 

R3D files therefore play on that oldest of human imbalances: fear vs. greed.

 

Fear comes from having been told that the R3D file is the equivalent of a film negative. You’ve heard what they used to do with those negatives, and what would happen if there was a scratch in any of them. You give yourself the goal of preserving the content of that negative at all costs.

But you are only partially right; you must preserve the negative, but at the least possible cost.

 

Greed happens when you attempt your first greenscreen key of a well shot R3D file. Yes, it can be a breathtaking experience, pulling the most effortless and beautiful key you’ve ever seen. Its siren song calls out to you, telling you to preserve those beautiful semi-translucent wisps of hair as if your life depended on it.

 

And then you see the result in D1, and you want to scale it 500% (possible!) so everyone can see what’s really there.

My role here is to talk you down off the ledge when your 15th render has failed because of memory errors. Here is what you must do: you must losslessly and effortlessly convert that R3D file down to whatever format you can actually use.

What did I just say? That’s right, you must throw away data that you can not and will not use. But - and this will make you feel better - you must also not throw away any data that you actually can use.

 

Right now it’s almost as if you must hire a consultant to tell the difference, which means that full exploration of this topic will have to happen in a future article which I may or may not entitle, “Convert an R3D with Confidence.”

 

For now, I’ll give you an outline. Yes, I know full well you’re not going to trust your half-million dollar commercial to an outline, but see if there’s anything here you can use: your source R3D file can be converted to several formats that are still better than your delivery format, and effectively lossless.

Your best bet in my opinion is to choose either 16 bit linear TIFF or 10 bit log DPX conversion and write image sequences; either format has more headroom than RED’s 12 bit linear sensor. You also have the option of QuickTime in whatever format suits your post-production pipeline - say, 4:4:4 10 bit Blackmagic HD, which will still key like a dream if well shot and won’t give up much in the way of color fidelity despite being a linear format.

 

“Uh,” you say, “but my situation isn’t that simple. The DP was like a drunken sailor, mixing in 50 fps 3K and 120 fps 2K footage of shattering glass and water droplets, framing wide on the 4K so that we would only use a portion of the shot.”  In that case you must take extra care to maintain each of those formats without keeping data you don’t need, but the bottom line is simple: sooner or later, you won’t be in full 4K (or probably even 3K) and sooner is better than later when it comes to dragging files around. This doesn’t even include the fact that you may only need a two second clip out of a two-minute long take, and if you stick with R3D, you must drag all 2 minutes of footage with you everywhere you go.

Give tip 2 its due, but keeping in mind your target, convert to 2K, 1K or whatever matches or slightly exceeds output, and forget that the R3D was ever meant to be seen by anyone. It’s your secret weapon, and your renders will seriously be 10-20x faster.
 
The caveat is that if you’re simply cutting and outputting, no effects or color correction, you could conceivably stay in R3D all the way until it’s time to render output or hand it off. That still leaves a couple of potential concerns.

(Page 2 of 3 pages for this article  <  1 2 3 >)

That’s how I do it Zak. offline in whatever format I need, make repos at will and when the project is conformed they repo from 4K and the shots look great. This is usually finishing in 1080 so you scale way up. This is conforming on a Quantel eQ who can read the R3Ds natively so it works well.

Posted by Scott Simmons  on  03/19  at  08:58 PM

---

>you can even create a color look in REDAlert or >Scratch and transfer it to an SD card to load it >directly onto the camera as a preview LUT.

>Many people don’t know this, and so it is the >rare shoot on which this happens.

http://www.reduser.net/forum/showthread.php?t=24810

http://www.reduser.net/forum/showthread.php?t=25905

You can only save and load looks created in camera. You can re-name those look files on a computer but cannot load a .rsx or .rcc file from redalert or redcine. I can’t speak on for the Scratch software, because I don’t own it.

Red says these feature will happen in the future.

http://www.reduser.net/forum/showthread.php?t=10060

Post #3

Posted by .(JavaScript must be enabled to view this email address)  on  03/20  at  12:02 AM

---

“Take control of sharpness, noise & more during conversion; don’t leave these to tools that can’t work with them”

Not sure if this works too well with vfx plates. We require unsharpened images, untreated as far as possible (though of course we can’t work with raw un-debayerized). For us, things works much better if the tweaks necessary for making the RED images look nice and crisp on the big screen happen downstream from us, NOT during the conversion out of RED raw into dpx or whatever.

This is what we did for the vfx on upcoming feature “Tormented” and it all worked very well.

Posted by paddy  on  03/20  at  06:28 AM

---

I’m gonna test creating on camera looks this week and will post what I find - thanks Steve.

Paddy, I understand the age-old habit of lowest possible sharpening on VFX plates, but I’ve found that some artists rely on the downsample to a lower res to do the sharpening “automatically” and that seems wrong given that a Bayer image needs sharpening. Do you stay at 4K and sharpen at the very end?

mc

Posted by Mark Christiansen  on  03/23  at  10:34 PM

---

“age-old habit?” Hmm - well, not quite the phrase I’d choose, but anyway…

Given that sharpening is only ever a trick to give a greater subjective impression of detail, without actually providing any additional information, it is certainly best left to the end of the image chain if vfx are to be involved. We don’t want to have to emulate the sharpening artefacts alongside everything else we do! Also - sharpening affects the image in non-obvious ways. It can make motion tracking much harder. It can make keyed edges noisier, etc.

Regarding what you say about using the interpolation filter when downsizing to add sharpness - I always recommend a “box” filter as the best choice. No fake sharpness, no aliasing, no ringing edges. It’s the most “honest” downsize, if one can put it that way.

On Tormented, all the sharpening did happen at the end - in the DI realm, so I have no detailed knowledge of what they did. But the editor and director are both highly tech savvy ( http://www.4klondon.com/ ), and I have no doubt they made a good choice.

In the vfx area - we did take some plates in at 4k - for instance if we were going to be zooming into any plates and applying post camera moves.

But the majority of the show was done with RED plates exported to us at 2048x1556 digital anamorphic. We horizontally squeezed the 4k raw 2.35:1 image to fill a 2048x1556 4:3 frame. Thereby doubling the vertical resolution of a standard 2k export.

The images looked a little on the soft side, compared, for instance with the aggressively sharp and grainy images I have seen from the ARRI D21, or some film scans. But nevertheless - very good. And, as they were very clean and low noise, I’m sure the DI sharpening options at the end were very good.

I have no doubt the film will look extremely good on the big screen.

Posted by paddy  on  03/24  at  04:37 AM

---

If your interested in using/working with LUTs, and if you’re using Red you should be, have a look here:

http://www.lightillusion.com/usingluts.htm

Once you grasp this a lot of the ‘mumbo-jumbo’ vanishes.

And this may alos be of interest - just my views on working with Red:

http://www.lightillusion.com/redworkflowissues.htm

Steve

Posted by Steve Shaw  on  03/27  at  04:01 AM

---

Name:

Email:

Location:

URL:

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below: