Where to put Color Checker in Node tree?

What is the Right Order for Color Checker in Node tree? No option I can think of give desired results. Is anyone else shooting an R5 and able to get resolve to play nice with the Clog3?

Camera, Canon R5. Latest firmware. Clog 3

Resolve 17.3.1, Rec 709 timeline

Following this guide, I have a X-rite Color Checker Passport Video

Follow are my color settings

Scope settings


To begin with, the Resolve CST for Canon log 3 to 709 seems broken. I have never been happy with the results I get out of it.

Follows is just a CST node from Clog 3 to 709. Reds are crazy saturated and everything is the wrong hue;



Using the Color Checker node after the CST. Brings Cyan to where I would expect it, but everything else is all over the place



Swapping to Color Checker first (Clog 3 to Clog 3). Green and blue are off the ranch.



How about just the Checker (clog3 to 709) and no CST at all? Again under saturated and Green/Blue problems.



I would expect if just the Clog 3 to 709 CST was broken, that the Color Checker afterwards could 'fix' it. But that isn't the case. And even just the checker alone is still junk.

Not to mention the drastically different outcome with different node orders.

Looks like you’re at least in the neighborhood. If the percentages underneath the chip squares in the UI are all low (at least under 10%, but the lower, the better), then you know you’re in the ballpark.

The node should go early in the chain, because it will normalize anything that goes before it. I have seen Patrick Inhoffer at Mixing Light use a curves adjustment first, to align the input closer to Resolve’s expectations, but that was probably back in version 12 or so.

Be careful to compare apples to apples. I’d recommend using the same frame to test different node setups, and a controlled studio setup if you want to A/B different cameras with the same node tree. Test one variable at a time. Differences in lighting is enough to mess with your results; time of day, cloud cover and spectral differences between studio lights will register.

If someone is holding the card, be sure their body doesn’t cast any shadows anywhere on its face. A thumb too close to the bottom black or white square will shift the gamma considerably. If the long grey patches don’t make thin, horizontal lines on the waveform monitor, the chart isn’t lit evenly. Be careful not to catch any reflections on the glossy black strip.

Do lots of tests, pick a method that gives you good results, then try to keep it consistent. When it comes to real world usage, there’s no guarantee that you’ll get proper charts with every scene change or light setup, if you get them at all, so use them to learn, but don’t let them become a crutch.

Btw, it looks like some of the scopes in your screenshots are set to 2x zoom and some aren’t.


Sent from my iPad using Tapatalk

Be careful to compare apples to apples. I’d recommend using the same frame to test different node setups, and a controlled studio setup if you want to A/B different cameras with the same node tree.

Btw, it looks like some of the scopes in your screenshots are set to 2x zoom and some aren’t.

Thank you for your response.

These were all the same frame of the same clip. My intention with these three tests was to try and find what is best, but all three seem like crap, so I assume there is something incorrect with my methodology.

The vector scope is on 2x the entire time. With both the 75% and 100% boxes on. ( i never found a guide saying which of these we are supposed to target). It looks like the zoom changed but that is because the Resolve Clog3 CST is just so far in left field, it is essentially unusable with its absurd saturation.

before X-Rite

2.-vectorscope-rec.2020-pq-sooc-2.jpg (365.11 KiB) Viewed 8551 times

after X-Rite

4.-vectorscope-rec.-2020-pq-corrected-2.jpg (379.65 KiB) Viewed 8551 times

Where the saturation lands depends on how the scene’s exposed, and on the type of chart you’re using, so don’t expect the dots to fill the squares exactly, but large shifts in the angle of a particular color versus the others is a clue that something’s not right.

The CST node is BMD’s best attempt to reverse engineer the color science of various manufacturers’ cameras. Since sensors vary between camera models (even from the same manufacturer), some combinations are more accurate than others.


Sent from my iPad using Tapatalk

2.-vectorscope-rec.2020-pq-sooc-2.jpg

4.-vectorscope-rec.-2020-pq-corrected-2.jpg

Thank you Jon, but I am sorry I don't understand what these two images are trying to show.

Also, the center blob on the vectorscope indicates an overall WB shift. In your first example, there’s significant shift.


Sent from my iPad using Tapatalk

Where the saturation lands depends on how the scene’s exposed, and on the type of chart you’re using, so don’t expect the dots to fill the squares exactly, but large shifts in the angle of a particular color versus the others is a clue that something’s not right.

Is that not the reason Resolve has the various test charts in the drop downs? Black Magic would have got the certifications from X-rite for what the 'Red' chip is on that specific model and then the tool can adjust the hue accordingly?

And even just the checker alone is still junk.

Just because I don't understand how to use a tool does not make it junk.

Also, the center blob on the vectorscope indicates an overall WB shift. In your first example, there’s significant shift.

Correct.

The R5's Whitebalace tool in camera was obviously designed by a sadist, you need to take a still, switch to movie mode and direct it to that still. Complete lunacy. So it is best guess in the field and fix in post.

None of the guides for the color checker mention if they expect you to correct white balance before or after the color checker node. It was my expectation that the color checker node would do it itself, it has white chips to look at.

I did try using the WB picker before/After the color checker and still never got results I was happy with. I didn't want to post all 9 options here.

But if someone knows the Right way I will certainly test.

The R5's Whitebalace tool in camera was obviously designed by a sadist, you need to take a still, switch to movie mode and direct it to that still. Complete lunacy.

That's why I never bought the R5.

And even just the checker alone is still junk.

Just because I don't understand how to use a tool does not make it junk.

Sorry, I meant the result I got from it are junk. That is undeniable. Not the tool itself.

I am here to hope to learn to use the tool correctly.

And even just the checker alone is still junk.

Just because I don't understand how to use a tool does not make it junk.

Sorry, I meant the result I got from it are junk. That is undeniable. Not the tool itself.

I am here to hope to learn to use the tool correctly.

Cool. I recommend doing it manually like so:

https://daejeonchronicles.com/2021/03/0 ... 7-part-ii/

Cool. I recommend doing it manually like so:

https://daejeonchronicles.com/2021/03/0 ... 7-part-ii/

Thank you, I have done that way, and obviously you get a perfect match on the vector scope. And it isn't actually that much longer than there color checker tool.

I guess I was looking for someone to say 'the color checker in resolve is broken in Resolve, you have to do it manually' to confirm the madness I am experiencing.

Cool. I recommend doing it manually like so:

https://daejeonchronicles.com/2021/03/0 ... 7-part-ii/

Thank you, I have done that way, and obviously you get a perfect match on the vector scope. And it isn't actually that much longer than there color checker tool.

I guess I was looking for someone to say 'the color checker in resolve is broken in Resolve, you have to do it manually' to confirm the madness I am experiencing.

I was in the same boat as you. The auto function never worked for me once. Colors were always wacky. hehe

Cool. I recommend doing it manually like so:

https://daejeonchronicles.com/2021/03/0 ... 7-part-ii/

Jon, Looking at your guide, Is the footage is log out of camera? Where do you do your color space transform? Before or after the custom curves?

Cool. I recommend doing it manually like so:

https://daejeonchronicles.com/2021/03/0 ... 7-part-ii/

Jon, Looking at your guide, Is the footage is log out of camera? Where do you do your color space transform? Before or after the custom curves?

Nowadays, my clips are ProRes 4444 transcoded from ProRes RAW in Compressor.

I use RCM, first node is WB, second node is X-Rite. When I'm finished, I put a noise reduction node at the beginning of my node tree. In his own tutorial, Matthew Allard writes:

The ColorChecker chart can also be used with Log recorded footage, although it doesn’t necessarily give you the best results from my experience. I found if you have Log material I got better results from first adding a LUT to the image and then doing the colour match.

Nowadays, my clips are ProRes 4444 transcoded from ProRes RAW in Compressor.

I use RCM, first node is WB, second node is X-Rite.

Is Compressor doing a Transform to linear there?

I gave up on using RCM as it is incompatible with fusion, so I use a Log-DWG CST at beginning of grade and DWG-709 CST at end.

But I expect it is functionally equivalent to RCM put the color checker after the first CST then?

Nowadays, my clips are ProRes 4444 transcoded from ProRes RAW in Compressor.

I use RCM, first node is WB, second node is X-Rite.

Is Compressor doing a Transform to linear there?

I gave up on using RCM as it is incompatible with fusion, so I use a Log-DWG CST at beginning of grade and DWG-709 CST at end.

But I expect it is functionally equivalent to RCM put the color checker after the first CST then?

Yes, Compressor is doing the transform to linear. Toward the bottom of the page, you can see the settings I'm using for Compressor.

https://daejeonchronicles.com/2021/04/3 ... rt-iv-raw/