Logarist Color Correction for DaVinci Resolve, Vegas Pro, Final Cut Pro X

Balazer,

I sent you a message with a link to the material you requested. Let me know if this stuff worked for you. Thanks.
 
I tried Logarist with some Sony FS7 footage. I liked how it made adjustments of exposure and white balance linear, definitely a quick way of working although I'd say that I don't find grading FS7's Slog3 particularly difficult "traditional" Logarist-less way. One thing I noticed unfortunately is a much increased appearance of banding and blockiness in the footage graded this way after it is exported out, particularly in shadow areas. It's been said that even 10 bits are kind of not enough or barely enough for 14 stops of latitude and Slog3- it's short comings are particularly critical in the shadows range which is starved of bits of information (mathematically). It get's even worse when the signal is linear as in the case of FS7 raw (where even 12 bits are not enough). My worry is that Loarists's first step which makes the footage even flatter contributes to that - my understanding is that the second of the Logarist LUT's refers to the levels that the first of Logarist's pair of LUT's creates (am I correct?). If I'm wrong, what could be the reason? If I'm right, I can't even imagine what happens in case of an 8 bit footage.
 
Hi, broughtonfilm. The banding and blockiness you're seeing were already present in the source footage, and aren't being made any worse by the processing. Logarist operates entirely in the high precision 32-bit floating point space, so it doesn't matter how "flat" the color space is. The log space is a faithful representation of the source. In fact the shadows of a log space are more spread out than in a regular camera or display color space.

What really matters for the appearance of that banding and blockiness is how those bottom stops are mapped into the output. What you want is for those bottom few stops to be mapped very low in the output, which means they'll be compressed to near black and have very low contrast. Your previous workflow was probably already doing that. You'll need to make sure that the same thing happens when you work in Logarist. Normally it's as simple as setting the exposure correctly in the camera and applying the appropriate exposure compensation afterwards. I haven't used the FS7, but from working with other S-Log cameras I know that typically you need to overexpose by as much as one to two stops, and then compensate in post. Of course you can also use a Shadow adjustment tool or a curve to give the shadows whatever specific treatment you want. Be careful too when reducing the contrast: that brings the shadows up.

If you want more help, please upload a short clip in S-Log 3 and some still frames of your corrected output. I can take a whack at it and try to figure out what went wrong.
 
balazer said:
Hi, broughtonfilm. The banding and blockiness you're seeing were already present in the source footage, and aren't being made any worse by the processing. Logarist operates entirely in the high precision 32-bit floating point space, so it doesn't matter how "flat" the color space is. The log space is a faithful representation of the source. In fact the shadows of a log space are more spread out than in a regular camera or display color space.

What really matters for the appearance of that banding and blockiness is how those bottom stops are mapped into the output. What you want is for those bottom few stops to be mapped very low in the output, which means they'll be compressed to near black and have very low contrast. Your previous workflow was probably already doing that. You'll need to make sure that the same thing happens when you work in Logarist. Normally it's as simple as setting the exposure correctly in the camera and applying the appropriate exposure compensation afterwards. I haven't used the FS7, but from working with other S-Log cameras I know that typically you need to overexpose by as much as one to two stops, and then compensate in post. Of course you can also use a Shadow adjustment tool or a curve to give the shadows whatever specific treatment you want. Be careful too when reducing the contrast: that brings the shadows up.

If you want more help, please upload a short clip in S-Log 3 and some still frames of your corrected output. I can take a whack at it and try to figure out what went wrong.
Click to expand...

It is not about how the shadows are spread out in log space but with how many bits these luminance values are described ( In any log signal the shadows are compromised when it comes to it). I never said that Logarist introduced the banding but that it made it more apparent. I'm describing my experience with Logarist based on actual use - I compared the output created with it with an output which I purposely matched with it in terms of the grade but created without it and the one made with Logarist has banding and emphasized macro blocking.
 
From the website:

"Logarist enables the same types of corrections available with raw images in Adobe Lightroom, but without the need to shoot in raw, and right inside your video editing application. "


For those who believe it I have a bridge to sell!
 
broughtonfilm said:
It is not about how the shadows are spread out in log space but with how many bits these luminance values are described ( In any log signal the shadows are compromised when it comes to it). I never said that Logarist introduced the banding but that it made it more apparent. I'm describing my experience with Logarist based on actual use - I compared the output created with it with an output which I purposely matched with it in terms of the grade but created without it and the one made with Logarist has banding and emphasized macro blocking.
Click to expand...
Yup, I understand, and I'm telling you that the appearance of the banding and blockiness has only to do with how much contrast you're giving them, regardless of which color correction approach you use. Give them low contrast and the problem goes away. Please post samples if you'd like to understand more.
 
Cary Knoop said:
From the website:

"Logarist enables the same types of corrections available with raw images in Adobe Lightroom, but without the need to shoot in raw, and right inside your video editing application. "


For those who believe it I have a bridge to sell!
Click to expand...
Cary, I'm not saying that compressed video offers the same quality and latitude for adjustments as raw. I'm only saying that the types of adjustments (how they work and what they do to the colors) is the same as with raw images. It's just basic color science and color management, same as used in ACES and in raw image processing.

I invite you to try the system before you criticize. It's free, after all.
 
balazer said:
Yup, I understand, and I'm telling you that the appearance of the banding and blockiness has only to do with how much contrast you're giving them, regardless of which color correction approach you use. Give them low contrast and the problem goes away. Please post samples if you'd like to understand more.
Click to expand...

What are you talking about? I'm telling you that footage exported with and without Logarist is the same in terms of contrast, colors, saturation, etc. - I matched it in DaVinci for the comparison purpose - the only difference is that the one created with Logarist shows banding and blocky artifacts and the other one does not. I'll post screen grabs of what I'm seeing later.
 
balazer said:
Cary, I'm not saying that compressed video offers the same quality and latitude for adjustments as raw. I'm only saying that the types of adjustments (how they work and what they do to the colors) is the same as with raw images. It's just basic color science and color management, same as used in ACES and in raw image processing.

I invite you to try the system before you criticize. It's free, after all.
Click to expand...
I understand what the system is trying to do and that is great, I am not criticizing that.

I think by staying away from "marketing LUT hype" and making those (confusing) claims about RAW images or comparing it to ACES you will actually increase the credibility of your system.

3D LUTs do have limitations however because values are interpolated. 'Stacking' LUTs upon LUTs will eventually degrade your image quality due to this interpolation.

ACES does not use LUTs it uses transforms which are far more accurate than LUTs.
 
Last edited:
broughtonfilm said:
What are you talking about? I'm telling you that footage exported with and without Logarist is the same in terms of contrast, colors, saturation, etc. - I matched it in DaVinci for the comparison purpose - the only difference is that the one created with Logarist shows banding and blocky artifacts and the other one does not. I'll post screen grabs of what I'm seeing later.
Click to expand...
Then maybe there is a problem with one of my LUTs. I would need sample files in S-Log to be able to debug that.
 
Cary Knoop said:
I understand what the system is trying to do and that is great, I am not criticizing that.

I think by staying away from "marketing LUT hype" and making those (confusing) claims about RAW images or comparing it to ACES you will actually increase the credibility of your system.

3D LUTs do have limitations however because values are interpolated. 'Stacking' LUTs upon LUTs will eventually degrade your image quality due to this interpolation.

ACES does not use LUTs it uses transforms which are far more accurate than LUTs.
Click to expand...
ACES is just a framework. You can do ACES with LUTs, CTLs, OpenColorIO, or hard coded transforms. Sure, LUTs have limited precision. LUTs are a compromise, to make the system fast and to bring it to more platforms. The color inaccuracies caused by the sequential application of two LUTs are pretty small and not as bad as you might expect. It's something I've tested extensively and compared to more precise transforms, and the trade-off for me is very clear: the speed and convenience outweigh the inaccuracies.
 
Balazer, just want to say thanks for your Logarist Color Correction and making it available to the community (at no charge I might add). Pleased with the images it creates from my GH4 footage and very easy to use when you get the hang of it and follow your very clear instructions.
 
Thanks, dr.eagle. Glad to hear it.


More about LUTs:

I've gone to a lot of trouble to make these LUTs as precise and accurate as I could. These are not like most of the LUTs you find available for download. These have a lot more points. I've tuned the LUT domains to cover just the color value ranges needed. I've paid particular attention to the behavior of the LUTs at the boundaries, extending the curves beyond the boundaries and not clipping, so that color values near the boundaries are as accurate as possible. To test LUT accuracy I apply a LUT and then the inverse LUT, and compare the output to the input. In all cases the color differences were imperceptible. You'll only see the differences in scopes, and even then the differences are very small. You can make the same test with my LUTs by round tripping from BT.709 to Logarist and back.

The conventional wisdom is that LUTs aren't precise enough for things like gamut transformations and sequential application. But I've found they are really just fine, if you are careful about the construction of the LUT.
 
To follow up on my previous post, I tried to match the clips as close as possible to present for comparison. I can't do it easily so the comparison would not be fair (mostly due to the fact that colors have slightly different hues so matching them requires more than a basic grade with color wheels and curves). I need to shoot a test using a color chart and will post it then.

I too wanted to say thank you for all the work that went towards creating Logarist and for sharing it with us.
 
Thanks, broughtonfilm, I appreciate it. I would like to try to solve whatever problems I can. I don't have an S-Log 3 camera, so maybe there is something going on there that I don't know about.
 
I rewatched your instructions and read your reply above again where you say: "To test LUT accuracy I apply a LUT and then the inverse LUT, and compare the output to the input. In all cases the color differences were imperceptible. You'll only see the differences in scopes, and even then the differences are very small. You can make the same test with my LUTs by round tripping from BT.709 to Logarist and back.
"
I missed that that the input before Logarist and output after inverse LUT are meant to look the same. I must be doing something wrong because when I follow your instructions for Davinci and use Logarist with S-log3 footage, after I apply the inverse LUT it does not bring the footage back to what the input was, instead it makes it look as if it was de-loged / rec 709-ish in terms of colors and contrast but over the top (with oversaturated colors and crushed blacks). I'm confused and until now I thought that it was meant to look this way, especially that with couple of tweaks I was able to get it to where I'd normally want it. But you say that "I apply a LUT and then the inverse LUT, and compare the output to the input" suggesting that what I should be expecting after the inverse LUT is applied is an S-Log3 looking output because that's what the input was. Am I making any sense?
 
Sorry, I think I added some confusion with my comment about testing LUT accuracy. What I meant was that if you apply these two LUTs in this sequence:

  1. BT.709 to Logarist
  2. Logarist to BT.709

with no corrections in between, the output should be almost the same as the input. The second LUT in that sequence is the inverse of the first, mathematically speaking.

I don't have an inverse of any of the S-Log3 LUTs, so you can't do the same test for S-Log3. No, I don't mean that you should get an S-Log3-like output.

You should be applying LUTs in this sequence:

  1. S-Log3 to Logarist
  2. Logarist to BT.709

with corrections in between. The output should look normal, not over the top. You probably will need some exposure compensation. But if the white balance was pretty close in the camera, you could get away without applying any white balance correction in Logarist. And most clips will look alright with no contrast adjustment, though most clips benefit from some. Generally you shouldn't need to do any saturation adjustment. You said it looked oversaturated. That suggests some mismatch between the gamut selected in the camera and the gamut of the input transform. Will correspond more over PM.
 
i downloaded and tried, a few notes..
- i'm working colorist
- i'm LUT phobic, most are pure crap, with tehedrial interpolation in Baselight or Resolve the LUT's show about as clean maths as any i have seen, a bit rougher as expected in Avid tho
- 99.5 % of the time i work in ACES, when working in Baselight and Nucoda i work mainly in the flim gradeing tools; exposure, contrast. pivot, saturation.. in Resolve i tend to use L*a*b as the filmstyle gradeing tool are pretty suboptimal to the point of being counter productive..
- i said all that so you'll kown that i'm very familiar with the general workflow and tools you are promoting... and far prefer using them.

- a shot from an FS7 / Slog3 in ACES with the 1.03 Slog IDT, and a 709 ODT does not match when i switch to display refered and use the S-Log3 to Logarist & Logarist to BT.709, i used Avid MC 8.7 for this, re-did in Resolve with identical resualts

- I see a workable starting point to grade from IMHO, and filmstyle gradeing tools are more workable in an NLE with this approach, unfortuantly every NLE's implemenatation of thse tools is even more sub-optimal than Resolve's, and i'm not aware of any that have a fully workable tool set, i'm very spoiled by Baselight and Nucoda's filmstyle color tools

Good to have the option for folks who want to take the time to figure out what these tools can bring to the table, thanks for shareing
 
You must log in or register to reply here.
Back
Top