Color space question

I'm working with 8 bit Sony HLG3. Sorry, lol. Don't think it matters for my question....

I prefer an unmanaged (project level) workflow. However, I want to take advantage of Davinci Wide Gamut/Intermediate at the timeline level, so I set that option at the project level. Project output is 709/2.4.

In my mind, that meant my node tree started with an IDT (sony to DavinciWG), then processed, then DavinciWG out to Rec709. I've seen others do that, too.

The problem is that it stretches out the waveform massively, to where you blow out some channels, especially red. In fact, in general everything leads heavy red channel with this method.

When I change to a single CST from Sony to 709, and then grade in front of it, these exaggerated results go away. So my brain and my eye aren't matching up. What am I doing wrong? Also, am I getting DWG aware tools in the timeline with that setup?

Could you share a screenshot of your 2 piece CST settings and your single CST?

dual cst output.PNG (259.59 KiB) Viewed 3549 times

dual cst input.PNG (260.22 KiB) Viewed 3549 times

single cst.PNG (169.37 KiB) Viewed 3549 times

Top is dual cst output, middle is dual cst input, bottom is single cst.

Could you share a screenshot of your 2 piece CST settings and your single CST?

There you go.

I guess the problem lies in the OOTF settings.

HLG has a different OOTF than Rec709 whoch makes talking about the process a bit confusing.

to compare both results do the following:
1. deactivate all your grading nodes except CSTs
2. in the dual CST setup disable OOTFs from HLG to DWG
3. Enable forward OOTD from DWG to Rec709
4. in the single node setup enable forward OOTF from HLG to Rec709

both images now should be nearly identical (there is a slight difference in brigtness, my guess is it has to do with the adaptation slider under tone mapping)

does the footage look the way you expect it to?

I guess the problem lies in the OOTF settings.

HLG has a different OOTF than Rec709 whoch makes talking about the process a bit confusing.

to compare both results do the following:
1. deactivate all your grading nodes except CSTs
2. in the dual CST setup disable OOTFs from HLG to DWG
3. Enable forward OOTD from DWG to Rec709
4. in the single node setup enable forward OOTF from HLG to Rec709

both images now should be nearly identical (there is a slight difference in brigtness, my guess is it has to do with the adaptation slider under tone mapping)

does the footage look the way you expect it to?

Thanks for your input!

No, that makes things worse. On top of the red channel still being blown through the roof (although the mid tones drop levels), when I disable OOTF's on the IDT, it crushes the lows into oblivion.

This brings up another question:

Do I even need or want an IDT node? It appears that each node's color space defaults to "use timeline", so since my timeline is DWG, shouldn't the color space aware tools work correctly? Is there any real reason to convert HLG into DWG? It's not as if you're adding data to the HLG signal...

Usually yes it's important to convert to your timeline color space in order to make the tools behave as expected.

the thing is, HLG3 is not a curve that is defined in the resolve CST. Resolve uses the official HLG curve as defined by the ITU Rec.2100. Sony seemed to modify that curve and call it HLG3, which makes it a blackbox.

Just to make sure, you are using YRGB unmanaged, only tagging the timeline color space and output color space, right? There's no transform happening after that whatsoever?

Usually yes it's important to convert to your timeline color space in order to make the tools behave as expected.

the thing is, HLG3 is not a curve that is defined in the resolve CST. Resolve uses the official HLG curve as defined by the ITU Rec.2100. Sony seemed to modify that curve and call it HLG3, which makes it a blackbox.

Just to make sure, you are using YRGB unmanaged, only tagging the timeline color space and output color space, right? There's no transform happening after that whatsoever?

Correct.

If I right click any node and look at it's color space, it defaults to "use timeline". Shouldn't that mean the color space aware tools are working in DWG?

Anyone else have ideas?

If I right click any node and look at it's color space, it defaults to "use timeline". Shouldn't that mean the color space aware tools are working in DWG?

Think of it this way. The HDR for example works in linear (at least the global wheel). So it does a transform from XX to linear, does math and goes back from linear to XX. In order to behave correctly it has to know what XX is. That's why you can set that XX in the three dot menu at the top of the tool. To make it easier, you set it to "use timeline" and define the timeline color space, instead of setting XX to the right color space all the time.

Anyway, let's get back to the topic. The way you set up the nodetree actually looks fine. can you show us an image of what the problem is?

If I right click any node and look at it's color space, it defaults to "use timeline". Shouldn't that mean the color space aware tools are working in DWG?

Think of it this way. The HDR for example works in linear (at least the global wheel). So it does a transform from XX to linear, does math and goes back from linear to XX. In order to behave correctly it has to know what XX is. That's why you can set that XX in the three dot menu at the top of the tool. To make it easier, you set it to "use timeline" and define the timeline color space, instead of setting XX to the right color space all the time.

Anyway, let's get back to the topic. The way you set up the nodetree actually looks fine. can you show us an image of what the problem is?

I'm sorry, but your answer confuses me more, lol. Am I correct, or not? In other words, if my timeline is set to DWG, and the color space aware tools always working in DWG unless I change the node to something else? Thanks for your patience.

Yes, I will post a couple pics to demonstrate...

Forgive me, but you've got your 8-bit(!) footage (what input color space?), you want to work in DWG (timeline) and you're outputting rec. 709. That's 3 settings. Why not just enter them into RCM and be done with it?

What is the attraction of using multiple transforms? What are you gaining, even forgetting the limited potential (and fragility?) of the 8-bit footage?

Forgive me, but you've got your 8-bit(!) footage (what input color space?), you want to work in DWG (timeline) and you're outputting rec. 709. That's 3 settings. Why not just enter them into RCM and be done with it?

What is the attraction of using multiple transforms? What are you gaining, even forgetting the limited potential (and fragility?) of the 8-bit footage?

Some people still work in 8 bit lmfao, jesus. Wow.

I want to control where conversions are done, which can't be done with a color-managed workflow. But thanks for your "answer" lmfao again.

cst only.PNG (788.34 KiB) Viewed 3416 times

cst only - red down.PNG (771.45 KiB) Viewed 3416 times

Here's CST at end only, and then pulling red highlights down.

Forgive me, but you've got your 8-bit(!) footage (what input color space?), you want to work in DWG (timeline) and you're outputting rec. 709. That's 3 settings. Why not just enter them into RCM and be done with it?

What is the attraction of using multiple transforms? What are you gaining, even forgetting the limited potential (and fragility?) of the 8-bit footage?

Some people still work in 8 bit lmfao, jesus. Wow.

Resolve's film emulations want to be fed 709, for one. Other reasons as well. But thanks for your "answer" lmfao again.

Resolve's film emulations do not expect 709!! they expect Cineon Log. for the gamut you can use rec709, yes but the gamma should be set to Cineon

Forgive me, but you've got your 8-bit(!) footage (what input color space?), you want to work in DWG (timeline) and you're outputting rec. 709. That's 3 settings. Why not just enter them into RCM and be done with it?

What is the attraction of using multiple transforms? What are you gaining, even forgetting the limited potential (and fragility?) of the 8-bit footage?

Some people still work in 8 bit lmfao, jesus. Wow.

Resolve's film emulations want to be fed 709, for one. Other reasons as well. But thanks for your "answer" lmfao again.

Resolve's film emulations do not expect 709!! they expect Cineon Log. for the gamut you can use rec709, yes but the gamma should be set to Cineon

Yes, Cineon then to 709. Not to DWG. Anyway, you're not helping answer my questions, you're trying to get me to change workflow. Not helpful and off topic.

cst only.PNG

cst only - red down.PNG

Here's CST at end only, and then pulling red highlights down.

you were right, my answers were kinda confusing. I wanna make sure I'm explaining it well.

only looking at the scopes doesn't really tell anything about the image. yes, the red channel is a bit hot, but depending on the image that could be expected behavior. How does the image itself look like?

idt odt no change.PNG (803.17 KiB) Viewed 3394 times

idt odt - red down linear.PNG (779.37 KiB) Viewed 3394 times

idt odt - red down curve.PNG (804.37 KiB) Viewed 3394 times

This is the behavior that I find strange. First is IDT/ODT, no changes. Second is a linear red channel pull down. Third is the curve I must apply to get back to close. Normal?

cst only.PNG

cst only - red down.PNG

Here's CST at end only, and then pulling red highlights down.

you were right, my answers were kinda confusing. I wanna make sure I'm explaining it well.

only looking at the scopes doesn't really tell anything about the image. yes, the red channel is a bit hot, but depending on the image that could be expected behavior. How does the image itself look like?

Got to step out, will post the actual images when I get back. It wouldn't let me post the whole screen, file too big. Thanks for helping!

wipe idt-odt on left.JPG (318.49 KiB) Viewed 3336 times

Here's a wipe and parade. IDT Rec2020/2100 to DWG then to ODT DWG to 709/2.4 on left, single CST from Rec2020/2100 to 709/2.4 on right. No corrections or changes on either. See how the IDT/ODT path is cranking everything, but especially the red channel? It also creates a halation type effect around really bright sources (not shown here).

Hey,

Sorry for the dumb question but in your initial screenshots where you show the setups the single CST variant has other nodes enabled as well. Was that still the case in your comparison test?

The other thing you could try is disabling the tonemapping when going from HLG to DWG, does that help the appearance? I don't see anything else out of the ordinary.

Hey,

Sorry for the dumb question but in your initial screenshots where you show the setups the single CST variant has other nodes enabled as well. Was that still the case in your comparison test?

The other thing you could try is disabling the tonemapping when going from HLG to DWG, does that help the appearance? I don't see anything else out of the ordinary.

I'm sorry, not your fault. I should have posted a pic without other nodes.

My comparisons have no adjustments, only CST's.

Disabling tone mapping as you've suggested does reign in some of the crazy contrast and boosted levels, but also make the image look very washed out and unnatural.

Anyone else? Wanting to solve this puzzle.

I mean it doesn't really look that wrong to me to be honest. sure the red channel is a bit hot, but that's where the color grading comes in. the overall white balance could be a bit cooled down, maybe decrease contrast, but those are creative decisions. technically it looks ok. Nothing wrong going on.

There is a difference between the two workflows because of the many transforms, tonemappings and OOTFs going on, but I don't see anything that wrong with the picture

I mean it doesn't really look that wrong to me to be honest. sure the red channel is a bit hot, but that's where the color grading comes in. the overall white balance could be a bit cooled down, maybe decrease contrast, but those are creative decisions. technically it looks ok. Nothing wrong going on.

There is a difference between the two workflows because of the many transforms, tonemappings and OOTFs going on, but I don't see anything that wrong with the picture

I very much appreciate your input! Trying to find a good overall system to save time.

To be fair I do think something is wrong because if you use a DRT to tone map media to display space none of the channels should 'clip' on the output. There is no roll off that I can see in that red channel it looks like it's going straight through the top of the display limit without any compression.

I don't have an HLG clip to test myself. If you could share a small non client test shot we could have a look at it.

To be fair I do think something is wrong because if you use a DRT to tone map media to display space none of the channels should 'clip' on the output. There is no roll off that I can see in that red channel it looks like it's going straight through the top of the display limit without any compression.

I don't have an HLG clip to test myself. If you could share a small non client test shot we could have a look at it.

I can get a clip, what's the best way to share it on here?

Something like a direct link to the file on your googledrive/dropbox if you have it, otherwise WeTransfer or something.

Tone mapping might be the issue. It wouldn't surprise me to be different using Davinci's specialized tone mapping on two separate color transforms instead of directly on one transform. Try turning it to none on all three nodes and see if it is then the same. If not, leave them at none and try changing the forward and inverse OOTFs as Sven suggested in his first post (I could also see that as an issue).

If you can get it to match, then you'll just have to deal with the issue in grading like Sven says, if that's the workflow you want to use. Although, you can try playing around with the adaptation slider or setting custom max inputs or outputs.

Tone mapping might be the issue. It wouldn't surprise me to be different using Davinci's specialized tone mapping on two separate color transforms instead of directly on one transform. Try turning it to none on all three nodes and see if it is then the same. If not, leave them at none and try changing the forward and inverse OOTFs as Sven suggested in his first post (I could also see that as an issue).

If you can get it to match, then you'll just have to deal with the issue in grading like Sven says, if that's the workflow you want to use. Although, you can try playing around with the adaptation slider or setting custom max inputs or outputs.

I've tried every combination. Nothing I do gets it to match. How strange.

To be fair I do think something is wrong because if you use a DRT to tone map media to display space none of the channels should 'clip' on the output.

Not entirely true. Very saturated or very bright areas can go into clipping even with a DRT applied. Note, we don't really know what the source looks like, and I assume it's one of the cheaper Sony cameras, so highlights might be pretty high anyway. Judging only the screenshot it looks fine to me, and if you can get some detail back in grading, even better.

You might try if saturation mapping helps, though.

To be fair I do think something is wrong because if you use a DRT to tone map media to display space none of the channels should 'clip' on the output.

Not entirely true. Very saturated or very bright areas can go into clipping even with a DRT applied. Note, we don't really know what the source looks like, and I assume it's one of the cheaper Sony cameras, so highlights might be pretty high anyway. Judging only the screenshot it looks fine to me, and if you can get some detail back in grading, even better.

You might try if saturation mapping helps, though.

Cameras are A7iii and a6600. I'll try to upload a clip today. Thanks again!

Cameras are A7iii and a6600. I'll try to upload a clip today.

I have A6600 and am interested in trying HLG in the future. Let me know if you need my help trying to recreate problem.

In the meantime check this thread viewtopic.php?f=21&t=178963 as it might help. Thread is not about HLG but in it Asim Siddiqui points to his articles about shooting and grading HLG.

Resolve's film emulations want to be fed 709, for one....

They want to be fed Cineon Film Log.

Good Luck

Resolve's film emulations want to be fed 709, for one....

They want to be fed Cineon Film Log.

Good Luck

Yes, I misspoke and corrected below that, Cineon to 709, but my reason for my preferred workflow is the same; I want to control where conversions are done, which can't be done with a color-managed workflow.

Cameras are A7iii and a6600. I'll try to upload a clip today.

I have A6600 and am interested in trying HLG in the future. Let me know if you need my help trying to recreate problem.

In the meantime check this thread viewtopic.php?f=21&t=178963 as it might help. Thread is not about HLG but in it Asim Siddiqui points to his articles about shooting and grading HLG.

Good find. Unfortunately, none of his suggestions for HLG seemed to make any difference, except turning off inverse OOTF, which I've already tried. Good read, though.

Unfortunately, none of his suggestions for HLG seemed to make any difference, except turning off inverse OOTF, which I've already tried.

So far the best thing I can find is to turn off OOTF and all tone mapping on both the IDT and ODT. It makes a very bright image, but at least nothing is clipping unusually. Still playing...