Compositing / Baselight Color Correction workflow to best utilize RAW?

[Todd M.]

Hey Guys,

I'm looking for some help here because I'm running up against a wall that I'm not sure there is a way around. We are shooting a video this week on Epic and there will be many VFX shots (mostly comps) in the finished product. We will be coloring on Baselight at a top facility here in LA and would obviously love to take advantage of the native .R3D capabilities of the system because we are going to be doing some heavy grading on the footage.

So here's the rub: with probably about 60% of the finished master being composite shots (sky replacements, FX plates, etc.) there does not seem to be any way to maintain the flexibility of RED Raw through to the Baselight finish. Our workflow will be:

Shoot Epic > Composite in AE CS5.5 > Color in Baselight​

Anybody have any tips on how best to run footage through this workflow? What would everyone suggest I transcode to in order to maintain the highest image fidelity and flexibility of for the color session?

Thanks for the input.

Todd

 

[Joseph Coleman]

Full Res 16bit openEXR. You can either transcode in Redcine or do your compositing on the original .R3D's inside of AE...

 

[Marc Wielage]

We will be coloring on Baselight at a top facility here in LA and would obviously love to take advantage of the native .R3D capabilities of the system because we are going to be doing some heavy grading on the footage.

Why don't you ask the facility what they'd prefer? Chances are, they have many years of experience at this.

I assume you're compositing in Nuke or Shake or some other VFX compositing program, and not trying to composite within Baselight itself (which can be done, but has some compromises). I have not heard of a heavy VFX project where they haven't done overall color correction prior to compositing, usually with DPX files, with some agreed-upon LUT applied. I have done sessions where I did two color-correction passes, one on the background plate, one on the foreground plate, and then the VFX artists create a holdout matte just by processing the green screen material separately and lining that up. There's a lot of ways to do this, but you the bottom line is that you can't color correct on foreground and background material separately without compromises -- unless you're actually doing the composite in the color-correction session, which again, is not ideal (but can be done). Multiple color-correction passes is what I've done in the past, sometimes literally shooting files out to the VFX people down the hall, depending on the workload and schedules.

Michael Cioni of Light Iron Digital has commented quite a bit about the workflow they used on Girl with the Dragon Tattoo, which had 1200 VFX shots, some of which were extremely complex. And all of that was 4K DPX files. As long as the colorist has the range to get what you need, and the temporary color correction applied is balanced and workable, you can get there.

 

[Hans von Sonntag]

Hi,

I'm shooting all my projects on Red and finish them either on Flame or Smoke. Equally to you, I have concerns regarding compromising quality down the post production process.

First of all, RAW does not hold any hidden information that a properly transcoded RGB file won't show. After editing I develop all consolidated shots to either RedLogFilm oder RedGamma2. Often redGamma2 is doing well enough and for Rec709 I find it easier to grade from. RedGamma2 holds whites very good, almost as good as RedLogFilm. Target files are 10Bit DPX. Going beyond 10 bit makes the system considerably slower. 12Bit is on the edge, 16 Bit is too much for Flame/Smoke and not necessary, IMHO. I cannot eyeball a difference and I don't see a difference when pulling a key, for instance. Things like proper lighting, fat exposure etc... are much more important than more Bit-depth beyond 10 Bit.

My advice is to transcode everything to DPX. I would let the colorist decide wether she/he prefers RedLogFilm or RedGamma2. Properly transcoded RGB DPX are as good as working off the R3Ds with the advantage of keeping things simple and the system snappy. I'm sure that the post house will recommend the same workflow.

Hans

 

[Dermot Shane]

Todd;

As you mentioned in your orignal post, there's a lot to be gained by using the r3d's directly in the grade system, to be able to change De-B settings in context is very useful in the real world, and it pays off in skin tones.. but i'm betting you already know this....

Depending on your timeframe... we work on Indiie features and MOW's mainly, so our sched's are written in weeks not days or hours - for those i send the r3d's to VFX (Nuke & Maya in my case) with my De-B settings, RMD, and a 3D LUT i generate from my base grade. I get back EXR at my sequence size, for the project i'm just finishing today it's DCi2K, so scaleing is done in Nuke.

I do have comp tools in my system, much like Pablo or Mystika so i can comp from elements, but normaly it's a better use of time in the gradeing suite to get the shots back as image seq's, at times it's better and not much slower to have elements returned and be able to do a final pass with the layers avb, you'll have to trade off time spent data wrangeling -VS- time spent masking elements that already exist as layers elsewhere... particles are a good example of this, i can vary the density, glows, colors with a real monitor.. a bigger deal when working in P3 than 709 due to most VFX having access to a fairly reliable 709 mon, but not the same story for P3 or a viewing enviroment to match P3... it does not take much of a gamma shift (say, to match a VFX shot to the rest of the seq after a creative choice is made after elements have been sent to VFX) and it will blow up the comp, to force the grade to be locked before VFX are in hand is not doing the project any favors

Maybe look at finishing / gradeing in hero box - Pablo, Mystika or DS all of whom have comping, native r3d and really good gradeing tools, might be a better bet than living with the limitations of Baselight / Nucoda / DaVinci.. and yeaaaa i know well the dedacated gradeing tools have strengths that the "hero box's" don't... mainly speed - but this seems to be a case where the "hero box" is an option worth exploring.

d

 

[M Most]

Maybe look at finishing / gradeing in hero box - Pablo, Mystika or DS all of whom have comping, native r3d and really good gradeing tools, might be a better bet than living with the limitations of Baselight / Nucoda / DaVinci.. and yeaaaa i know well the dedacated gradeing tools have strengths that the "hero box's" don't... mainly speed - but this seems to be a case where the "hero box" is an option worth exploring.

DI's aren't done by boxes. They're done by colorists. The capabilities of the boxes aren't worth much if you don't have a colorist who "gets" what you're trying to accomplish in terms of color, a toolset that he or she can use to accomplish that task, and a proper environment to do that job properly. The vast majority of "A" title DI's involve huge numbers of VFX shots, but the DI is not usually done on a "hero box." At Company 3, DI color work is done on Resolve, at EFilm it's done on a version of Lustre, at Colorworks it's done on Baselight, at Technicolor it's done on either Lustre or Resolve - you get the picture. And between those four companies, you've probably accounted for about 80% or more of the features you see in theaters, and perhaps 90% of all the "tentpole" features. DI's, like much of post production at the larger budget levels, are done in a departmental fashion, where the composites are done by compositors and the color is done by a colorist. Each has his or her own specific equipment for those tasks, and they are usually not the same thing, just as the skills involved are not always the same.

As for the question being asked, my experience is that VFX elements are not always pre-colored, as Marc was implying. As long as the images are kept in log space, viewing LUTs can be supplied to ensure some sort of consistency. It's only when the original photography is being seriously manipulated or distressed for a "look" that pre-coloring is really necessary. More often than not, keeping the images in a log format (RedlogFilm represents the most common of these, Cineon Log) will retain all information in the original image and yield enough flexibility that the returned VFX images can be matched to the original images in a pretty straightforward manner. DPX is the most common container for that because it can carry metadata for positive identification, and it is optimized pretty well for real time playback on most systems, assuming they've got fast enough storage. And I agree with Hans' feeling that 10 bits is enough for this use most of the time, provided the images are handled properly.

 

[Dermot Shane]

DI's aren't done by boxes. They're done by colorists. The capabilities of the boxes aren't worth much if you don't have a colorist who "gets" what you're trying to accomplish in terms of color, a toolset that he or she can use to accomplish that task, and a proper environment to do that job properly.

agreed..
now is it worth exploring weither the colorist he prefers is inside the comfort zone on any of the hero box's?

Assuming;
1) the hero box is set up to do accurate color work (most, but not all are)
2) He's happy with the colorist attached to the project already
3) the VFX are primarily CGI and can be delivered as elements in EXR

I'm thinking that it's a path worth exploring, not every project is going to fit any given toolset, even one as compentent as Baselight runs into finite limits eventualy.. just as there's reasons why hero box's don't do pipeline as effecently as a dedicated gradeing toolset.

d

 

[jake blackstone]

FilmLight's Nuke plugin is custom made for just such situations. The idea should work like this:
1. Send the shot to NUKE.
2. Open it with Baselight grade applied or not, your choice.
3. Do the composite and without rendering it, export it to Baselight.
4. Open it in Baselight and grade it, as you were inside the Nuke, all nodes, metadata, at full size, all intact.
The beauty of this is, that no color grading decisions are final until the very end. Oh, did I mention, that there is a BL plugin for FCP too? Personally, I'm hoping for MC6 plugin. Anyway, everything is done in metadata, nothing ever gets baked in until the very end.
Conceivably, one could open conformed Baselight timeline down the road in, say, 5 years, apply RedColor 5(?) and being able to regrade it all over again with new color science, including VFX shots...
I think, FilmLight may start shipping it at NAB time.