RedCode RAW vs RedLogFilm/12-bit ProRes4444 for color correction -- any actual tests?

[Erich Ocean]

So, I know that RAW is awesome. Especially compared to the alternative on DSRs: 8-bit JPEGs. That much makes sense, no need to discuss how much better RAW is in that comparison.

But what about RAW vs. RedLogFilm 12-bit ProRes4444? Clearly, the latter is a totally different beast from an 8-bit JPEG. Maybe RAW capture would never have even been invented if that was the alternative, because the actual improvement would have been so small as to have not been worth the trouble.

So here's what I want to know. Has anyone ever had RED footage that they could not get the grade they wanted to get from a RedLogFilm 12-bit ProRes4444 file that they *could* get the grade out of the RedCode RAW data. This is not "RAW is better" or "RAW always works", I want to know if someone has actually run across a situation where RAW "worked" and the RedLogFilm did not.

In other words, how much "extra" is there in the RAW that's not in the RedLogFilm version? Because if it's minor, or no one in the real world has ever had a RedLogFilm image they couldn't get the grade they wanted on, then I'll be much more inclined to just transcode my RAW footage to that and never think about the RAW again.

Anyone?

 

[Jeff Kilgroe]

12bit log ProRes4444 has a ton of information in it. As far as color range goes, there's not a whole lot less than the RAW master files. It's more an issue of what goes on in the conversion process from RAW into an RGB matrix. You lose precision each time you translate or convert. You also introduce other factors to affect image quality, such as recompression.

The theory behind RAW isn't just about the color information. It's about recording the image data in the most un-touched format possible. You then use that RAW camera data to create footage in your desired format. Thus keeping the decision of what format to use for editing, mastering, finishing, etc.. as a choice to be made in post -- where it can be decided what is best for the project at hand.

A great example of how onboard RAW recording can be better is to look at the Alexa. Many people rave about the ability to directly record to ProRes. Well, that's fine and dandy, if that works for you. If that gives you the quality you need. Recording ArriRAW from the Alexa is an expensive proposition with an external recorder. And that can often be a necessity if someone chooses to shoot that camera, but needs something more than the ProRes files or needs to work in a post pipeline that doesn't play well with ProRes -- such as most PC based workflows, Avid, Scratch. If you record to ProRes, you're already one generation away from the data the camera acquired. Then if you need to put it into a context of DNxHD or other format, you're another generation away. You can minimize much of the generational loss by moving into 32bit float space and/or going to uncompressed, but that's not always a practical choice.

Being able to deal directly with the RAW files just provides the most flexibility in many respects. Not just in terms of color/luma range and accuracy. Having software that can natively work with the RAW files throughout the post process, from editing to finish allows putting off the conversion from RAW until the point it makes the most sense and simplifies workflows by allowing to use the same files universally rather than creating alternative proxy files and re-linking original camera media later.

 

[Alexander Ibrahim]

The answer depends on your workflow.

If I do a grade before transcode then I have never had any problems getting the footage to go where I wanted it later.

Key point, it has to be a pretty good grade. Not too aggressive, because you want to preserve data, but you already must be taking the footage in the direction of the finish.

You can't just get the white balance "in the neighborhood" and throw a default look on everything.

Without a grade I have had issues, and needed to go back to the RAW and regrade/output.

There were quite a few shots I did under 17 that were unsatisfactory. They look a lot better when regraded in RedCine X recently.

So the short answer is stick with the RAW until you can do a good grade in RedCine or Storm as a minimum.

You will get better results if you keep the Raw in your workflow until late in the game and take it in to Color or Resolve.

 

[Frank Cueto]

Also, raw allows easier fixes (like kelvin/iso/flut/olpf/drx)

Its also nice to grade raw, render the 4k (that doesnt take up much space) and downconvert to hd/2K.

 

[Rainer Fritz]

12bit log ProRes4444 has a ton of information in it. As far as color range goes, there's not a whole lot less than the RAW master files. It's more an issue of what goes on in the conversion process from RAW into an RGB matrix. You lose precision each time you translate or convert. You also introduce other factors to affect image quality, such as recompression.

The theory behind RAW isn't just about the color information. It's about recording the image data in the most un-touched format possible. You then use that RAW camera data to create footage in your desired format. Thus keeping the decision of what format to use for editing, mastering, finishing, etc.. as a choice to be made in post -- where it can be decided what is best for the project at hand.

A great example of how onboard RAW recording can be better is to look at the Alexa. Many people rave about the ability to directly record to ProRes. Well, that's fine and dandy, if that works for you. If that gives you the quality you need. Recording ArriRAW from the Alexa is an expensive proposition with an external recorder. And that can often be a necessity if someone chooses to shoot that camera, but needs something more than the ProRes files or needs to work in a post pipeline that doesn't play well with ProRes -- such as most PC based workflows, Avid, Scratch. If you record to ProRes, you're already one generation away from the data the camera acquired. Then if you need to put it into a context of DNxHD or other format, you're another generation away. You can minimize much of the generational loss by moving into 32bit float space and/or going to uncompressed, but that's not always a practical choice.

Being able to deal directly with the RAW files just provides the most flexibility in many respects. Not just in terms of color/luma range and accuracy. Having software that can natively work with the RAW files throughout the post process, from editing to finish allows putting off the conversion from RAW until the point it makes the most sense and simplifies workflows by allowing to use the same files universally rather than creating alternative proxy files and re-linking original camera media later.

Great post Jeff ! It describes very simple what the power of red-raw is...

 

[Erich Ocean]

The answer depends on your workflow.

If I do a grade before transcode then I have never had any problems getting the footage to go where I wanted it later.

Key point, it has to be a pretty good grade. Not too aggressive, because you want to preserve data, but you already must be taking the footage in the direction of the finish.

Were you doing the initial "grade" to a RedLogFilm/ProRes4444 output? I can definitely see what you're talking about if you're baking in gamma. But RedLogFilm doesn't do that; it's basically like outputting to floating point linear (but more efficient), so all that's *really* baked in is the color matrix/white balance -- something that's very easy to do on set. (Graeme, if you're reading, please correct me if I'm wrong here...)

I don't have a horse in the race (or even know the answer, not having worked with RED footage before). For most of the usage I have, I need a live output in linear space, so RedColor2/RedLogFilm out the HD-SDI port is ideal on the Epic. I'll still record RAW (of course), but I got to thinking that the HD-SDI output with a Ki-Pro mini (yes, only 10-bit 4:2:2 at the moment, but in the future, I expect RED to have a 12-bit ProRes4444 module to match the Alexa) might not actually lose much, if anything, and then I can skip the transcode step or buying a RedRocket -- the Ki Pro mini is something like half the cost.

Anyway, thanks for the info. You at least were able to confirm that people are having to go back to the RAWs occasionally, especially with older firmware builds. Thanks!

 

[Jeff Kilgroe]

Also, raw allows easier fixes (like kelvin/iso/flut/olpf/drx)

Precisely. And that seems to be something I didn't touch on very well in my post. When we have RAW sensor data to work with, we can still set the mid grey (ISO), OLPF compensation settings and more before the initial translation into a defined color space. Once again, it's just another area where RAW provides more control, compared to having this all done in-camera and such settings being permanently baked into your recorded images.

 

[Gunleik Groven]

OK. Playin' the devil here...

1st: I agree with Jeff. Sorta. Generally...
Actually depends a bit on your tools and the skillsets of the colorist.

The main advantage of RAW - to me - is in the capture stage, because as colorscience and debayering techniques improve, so will your images. In that respect, I cannot see how a prores 444 capture could replace those advantages.
I have done some comparisons with a prores 4444 12 bit camera....

Grading from RAW vs uncompressed/prores4444 might be a slightly different scenario.

1.
You are dependant on a good "meta grade" whichever way you go. Not all colorists understand how that work. Thus you may easily get a better result from the pregraded redlogfilm material.

2.
Many tools are very well tuned for a cineon signal, and thus works better with that.

3.
I know you din't mention EXR, but we did an a/b between 10 bit DPX and open EXR. Really just a quick diff-matte test. And the extra signal in the EXR was significant.
Theoretically you should be able to carry that extra beef through in a RAW grade. Whether that is a bonus or not, depends a bit on how the meta-grade is done, though.

We are going from uncompressed RedLogFilm to RAW grade (in Resolve) for our next series, but will still keep the first-light step as a part of the conform and do EXR for VFX.

 

[Alexander Ibrahim]

Were you doing the initial "grade" to a RedLogFilm/ProRes4444 output? I can definitely see what you're talking about if you're baking in gamma. But RedLogFilm doesn't do that; it's basically like outputting to floating point linear (but more efficient), so all that's *really* baked in is the color matrix/white balance -- something that's very easy to do on set. (Graeme, if you're reading, please correct me if I'm wrong here...)

This isn't a function of RedLogFilm, but rather of ProRes.

Once you do an initial grade in RedCineX ... or whatever tool you like, then output it to a "traditional" video codec you leave RAW behind.

So, if you grade your material intelligently and output it to a good format preserving data, and I do consider ProRes4444 a good format, then you can proceed with a normal "video" workflow if you choose without fear that you'll lose something.

If you just do a simple correction of some sort that's very quick (and perhaps slightly automatic) You will get caught out a few times and have to go back to the RAW.

So, my definitive preference is to go through post with RAW until final output.

My second choice (and the choice I use most often in practice) is to grade the film, as best I can given the schedule, then output to a video codec and go through a normal post that way.

If you do a quick adjustment and output, you will get caught and have to go back to the RAW at some point in post ... but even that isn't so bad.

Red post is definitely getting easier, but I think we still have a few iterations of technology and software before it gets really smooth.

 

[alexx_hg]

Also, raw allows easier fixes (like kelvin/iso/flut/olpf/drx)

Exactly. Just think about a longform project (outdoor), which takes two or three different seasons. Adapting in RAW is a childplay compared to burnt in data.

IMHO, beside all RED´s Hardware, bringing RAW to motion pictures was the best invention what RED has done.

bg
alex

 

[Nick W]

Guys forgive me joining in here as a newbie. My background is stills and the RED workflow is still new to me. However, I do have a heap of experience with RAW.

I was in RAW in the early days (10 years ago +) and the key point is that RAW footage is like a fine wine - it keeps getting better as software evolves. As others have said, once those pixels have been de-bayered, even to a 16 bit or greater codec, everything is locked down and lossy. RAW is still new to video, but as software evolves, you will find that it is possible to bring out expressions in the image that cannot even be seen at this moment in time.

IMO the key to greater efficiencies are improved grading and editing tools that allow the complete edit and grade to be done in RAW, with the final output being the first generation of ProRes / ANother codec.

I have seen this evoloution in the stills world, to the point that we are now able to do much of the photoshop-type work to RAW images, before output. The lesons learned there should ensure that these developments happen much faster with moving image.

There are companies out there with sophisticated processing technology that can be licensed and in the sohort term that may be the way to jump forwards?

Nick

 

[Erich Ocean]

Okay, sounds like the verdict is that unless you can more or less hit close to the final grade for every shot, you'll need to go back to the RAW footage at some point.

Proving, once again, how f'ing cool it is to be able to shoot RAW motion pictures. :beer: