FS7: Catalyst Browse adjustments and LUTs look different in Prmeiere

Ben Scott

Ben Scott

Veteran
Looking at the scopes as well as the image, if I stick a clip in Catalyst Browse and export a LUT based on my adjustments and then apply that LUT to the footage in Premiere, the results are markedly different.

Is there a workaround for this, or is Catalyst Browse useless for anything other than transcoding the footage?
 
Ben,

Catalyst Browse takes the same approach as DaVinci Resolve, where the clip is imported as data levels, but then worked on as legal levels - so you are monitoring legal levels. LUTs are data in / legal out. Premiere, with the Lumetri plugin treats the entire workflow as data levels, displaying legal for monitoring on the fly. LUTs for Premiere / Speedgrade should be data in / data out.

My LUTCalc app can be used for reading in a Catalyst-generated LUT and then adjusting it, or you could stack your Catalyst LUT followed by a legal-to-data LUT in Premiere, but then that is an extra complication!

I've uploaded a new version of LUTCalc today, which can read a variety of LUT formats and output appropriate levels. The LUTCalc For Mac version can also now import 16-bit tiffs and pngs. Links:

LUTCalc Main Page (with online version)
LUTCalc For Mac releases page on GitHub

Ben
 
Hey matey,

Thanks for chiming in. I assumed that was the reason behind it but didn't know what to do about it.

Intend to get to grips with what sounds like awesome witchcraft in your app when I have a spare second but it sounds like in the short term sack of Catalyst, which to be honest isn't really a big deal.

I wonder if the new 2015 CC has simple controls to interpret video/data levels.
 
Let me know if it works!

You can also just use a levels filter and squash the image a bit to reduce its contrast and put it back to slog-like levels. Here's a reference image I put together that shows you what to adjust:

Untitled_zps4653c42e.jpg
 
Dennis, Ben,

Dennis' LUT looks to be pretty close to a straight legal to data conversion. One thing is that as with most LUTs it takes input value between 0 and 1.0. That means that if your corrections take values outside of legal range before applying the legal to data LUT, they will get clipped. Not really an issue for finishing up, but if you want to do any more tweaking after (or you are working to extended range)...

The cube format allows for scaled inputs - though the Adobe / Lumetri format is subtly different to the DaVinci format. I've quickly knocked up a couple of legal to data 1D LUTs in the two formats which will take input values from -25% to +200% of legal range, so hopefully no danger of clipping!

Link:
Legal to Data 1D conversion LUTs for Lumetri and DaVinci

Ben
 
How I imagine you would use this is import clip, add my LUT to correct levels, then after you apply your corrections or 3rd party LUTs. You would not really want to correct before the correction LUT or yes you can end up clipping data.

I haven't tested every NLE out there, but I do know for example that a version of FCP X I was testing will permanently clip the highlights in its internal levels scaling process, whereas for example in Davinci Resolve, even if you tell resolve to treat a data levels file as "legal levels" and it clips on the waveform monitor you can actually just reduce the gain or highlights, etc.. and the highlight data is still there. In the case of FCP X there was a proper plugin or work-around to being able to access those higher levels that inherently FCP X wants to clip.

I'm not too sure how Premiere handles it, but always test your NLE workflow first to know its limits.
 
I know its been stated before , but could you guys just briefly explain the difference between data in legal out vs data in data out?
Thanks lenny
 
That article is an ok primer but references data levels in terms of DI (or post production) workflow and what it means but not at all what "data levels" means in terms of a video recorded image.

So one thing that really keeps me up at night is whether or not Sony SLOG is true "data levels" like the article refers to or is it actually just an extended version of video levels which is also very often referred to as "full range" (improperly if you ask me) or otherwise referred to as "illegal" or sometimes just "extended".

Head over to the white paper on the new Slog3 v1.0 PDF and in their FAQ section they pose the question: "is slog3 full (extended) range or legal range?"

Shouldn't the question have been: "is slog3 data levels or legal range?"

The answer goes on to explain that slog is recorded as full range in XAVC, MPEG and HDCAM SR file. It then says that a data range is used from 0-1023 to utilize the full 10-bit code value.

I hate to call Sony liars but first thing first, all the codecs they listed are YCrBr and use REC709 space to store video. This is 64-940 and if you go into "extended" rang you can get up to 1019 but no higher because the last few bits are reserved for other things. Second if you import any slog shot clip into Davinci Resolve without touching any setting at all it treats it as a "video level" 64-940 clip by default and if you import a "true" RGB data levels clip (like EXR, or RGB file from photoshop) it automatically treats it as a "data levels" 0-1023 clip.

If slog really was "data levels" 0-1023 then Davinci would know it. What I suspect is we are just treating the slog in extended video signal mode as a "data levels" signal because they are almost identical in bit value - but to me they are not.

The only video formats I'm aware of that would be true 100% data levels (0-1023) in origin would be ProRes 4444, RAW, and maybe the SR codec in RGB.

What do you guys think? Am I over analyzing here? Should I stop thinking about this when I'm sleeping?


Levels keeping YOU up at night? Check out my video/data levels demystified article: http://www.hingsberg.com/index.php/2014/10/video-data-levels-demystified/
 
Last edited:
Thanks Liam, that was very clear.
It brings to mind a number of questions for me:
- Is it easy to see whether your NLE is working in data or legal levels or do you need to just know it ahead of time?

I'm still in good old FCP 7 which uses IRE measured waveform displays - do data numbers simply translate to IRE?: 255 = 109IRE and 235 =100IRE or is something more confusing going on?
In FCP7 I assume its working in data levels since if I shoot 109% whites they show up as 109% and I need to correct to get them down to 100. Is that correct?

IN Catalyst Browse the waveform is measuring IRE and only goes up to 100% and the image clips at that point. I assume that is what's meant by Ben's description of it as Data-In but working in Legal. If you transcode without color correction I assume it just keeps the data levels - right?

So I wonder if this has anything to do with something that has been confusing me recently:
I've been shooting a few interviews on the FS7 using extended LUT's (2%-109%) from LUTCalc (Data-IN Legal-Out as I think Ben suggests) and burning them in. I monitor in the field using a Flanders Scientific 17".

I've noticed that when the picture seems to look good to me the waveform looks a little low - Faces closer to 50% than my usual standard of 60%. However it looks hot if I go to 60 which is what I would normally do in 709. I get surprised in post though. When I get the pics into Catalyst Browse they look slightly underexposed. In fact the blacks also look higher than I expect them to.
I've never noticed this when shooting 709 and assume it's just something in my monitoring habits that have to adjust to the LUT's, but could it have anything to do with this data-legal business? I haven't had time to test this carefully yet.

Thanks for putting up with my less than stellar technical knowledge and spreading the 'light" around.

Lenny Levy
 
Last edited:
FCP7 or X is working in legal levels, if you bring highlights down they are still there like you said. I have seen NLE's clip this instead but maybe with 7 it works like Resolve and you just scale down what's actually there.

Catalyst should do the same thing but i'm not at my computer right now, it might show it as not going past 100% but if you reduce the highlights the data is there.

A for output from Catalyst I just tested it recently and the old version of Browse would always output a legal version scaled down from "data levels" (it's not real data levels but fine let's pretend it is) which is a problem if you wanted ProRes versions of your slog so it could head elsewhere for grading, but the new updates version of Browse can output with the full levels which is great.
 
Last edited:
You guys rock, you should open your own "(Sony) filmmaking school" :D It would be extremely successful!! Thanks for all that great information!!!!



Edit:
Question ;) When working with 12bit footage are all those ranges scaled accordingly? Also does nle's show 0-4095 range or it's stack on 0-1023? (as when working with 8bit footage scopes are with 0-1023 measurements).
 
Last edited:
alohype said:
You guys rock, you should open your own "(Sony) filmmaking school" :D It would be extremely successful!! Thanks for all that great information!!!!



Edit:
Question ;) When working with 12bit footage are all those ranges scaled accordingly? Also does nle's show 0-4095 range or it's stack on 0-1023? (as when working with 8bit footage scopes are with 0-1023 measurements).
Click to expand...


If Sony wants to pay me I would be happy. Right now I'm doing it for free without knowing it? :cheesy:


Answer ;) trick question actually. Normally bit depth has nothing to do with what levels are being used, unless of course you are talking about ProRes 4444 12-bit which is always "true", "real", 0-1023 RGB "data levels". But surely there are other 12-bit codes that will store video as "video levels" with "extended" (or aka "illegal") code values.

In resolve you can change you scope to read 0-4096 or whatever you want instead of 0-1023 but I'm not sure about other NLE's.
 
You must log in or register to reply here.
Back
Top