Blackmagic Forum

Hello,

My typical process for grading log footage of BMPC 4K or Micro Cinema DNG has been to adjust temp and tint at the first step - either in RAW panel or first node, and then after some minor adjustment to exposure, lift, gain - again on either raw panel or first 2 nodes, to use log color wheels and low cutoff knob to cool the shadows slightly (these cameras need it). Only then do I raise game to bring the image fully out of a log profile - I do this with the tone curve rather than a LUT.

the 3:1 DNG raw footage from Micro Cinema camera is first being interpreted using the RAW panel BMD color space, gamma, and color science. For prores HQ from the BMPC, I don't apply any LUT - just the process above.

Is this incorrect? When I've seen LUTs applied to de-log the footage, it tends to look off. Maybe i part because it was shot with bad temp and tint settings.

A colorist I was working with said he thought it would impossible to achieve a natural result without using LUTs as a first step.

Thanks for any help!

I won't say it's "incorrect". I do think you're making it more complicated than it needs to be.

Try these settings, see what you think.

I'll let the real colourists speak but I think the problem with adjusting while in log is you can't use your eyes. If you're happy working with the tools then I don't think it's wrong.

It wouldn't work for me but my eyes know more than I do. I can tell an image looks wrong but I can't always look at the scope and know. Make any sense?

It's not strictly wrong but various approaches come with various pros and cons. It comes down to preference and knowing what can go wrong. Especially now Resolve is slowly getting more color managed tools, approaching things non-managed can become problematic.

Here is a video on Baselight and using it in Telecine-style grading which is essentially what you're doing.
Their software has clean solutions to do hybrids but Resolve does not. You can convert a node internally into a new color space but not identify a node as a new color space.


I would say personally that the best approach is a scene-referred (log) grade with proper tagged timeline space (match grading space) and using LUTs/CST/RCM to do display conversion as last step, and if desired use non managed tools post display conversion only to do slight tweaks that are hard to achieve underneath display conversion/tone mapping.

Creating the display rendering itself through grading is futile imo because the available tools are not built to effortlessly create a predictable, consistent and robuust rendering of camera input data. There is also zero benefit to do that work manually.

Thanks, it will take me some time to digest all of this.

I do want to mention that although I have nodes that are operating in the log space, my monitoring is not in log - subsequent nodes make that conversion, and I'm monitoring at the end of the chain. So it's not the case that I'm adjusting without seeing the final result.

I thought that Davinci's log wheels were named as such because they were intended for use on log material.

Jim Simon wrote:I won't say it's "incorrect". I do think you're making it more complicated than it needs to be.

Try these settings, see what you think.

Rec 709 RCM.png

This is useful. Thank you.

The footage that is DNG 3:1 from Pocket 4K sometimes comes through color managed looking much too saturated. Seems a strange place to start. Color managed from micro cinema DNGs look good color managed. Color managed doesn't seem to do anything with prores HQ from Production camera 4k. I guess that would take a LUT.

Also I think I saw you can through color management onto a node rather than have it baked into the timeline? So that you can pick which clips to deploy it on... Anyone have an idea on that?

Thanks everyone for your help and please excuse my amateur questions.

The general idea in modern practical grading workflows is that you convert your various camera capture data into a single log grading space. All the corrective and creative grading happens there and after it the display rendering gets applied as a single transformation for all the graded clips, be it via the timeline level node tree, on group post clip level node tree or through RCM. This flow is in order to have consistent behaviour when grading various sources because you aren't switching the grading space every time you do so.

It doesn't matter if you apply this principle through RCM like picking RCM HDR preset or the non automatic DWG/Intermediate preset for example. Or totally manually with nodes. The main difference between node approach and RCM is that nodes allow you to do operations after the display conversion since it's just a node in the chain. You cannot do this in RCM without a 'workaround'.

When using RCM, Resolve uses embedded metadata to auto assign the correct input color space to non-raw media. If this isn't detected it will default to 'from project settings'. You have to select all your ProResHQ media and right-click -> Input Color Space and choose the correct one.

Raw media is automatically correctly managed. If footage looks too saturated from DNG raw files something else is not correctly set up.

shebbe wrote: You have to select all your ProResHQ media and right-click -> Input Color Space and choose the correct one.

The info posted to this thread has been useful to me. Based on these recomendations I'm trying the following workflow:
Timeline:
colorspace Davinci WG/intermediate
Output: Rec709 2.4

first node on each clip:
Colorspace transform
Input color and gamma: [manually match to camera]
Output color and gamma: use timeline

last node on each clip:
Input color and gamma: use timeline
Output color: Rec 709
Output Gamma: Gamma 2.4

Additionally, I have my monitor set to Rec 709, gamma 2.4. I had it before on sRGB, not seeing a major difference there.

One question I have - are my grade nodes between the two transforms essentially working on log encoded images? Or is it that the WG/intermediate color space is is too wide gamut for my monitor, so it looks drab unless the end transform to rec709 is engaged?

I also have the following question. Why is it necessary to transform to rec709 on a final node, when the timeline settings already specify an output of rec709? Am I making a mistake by doing that transform twice?

Benjamin de Menil wrote:Pocket 4K sometimes comes through color managed looking much too saturated.

Without any nodes? Even in the Source Viewer?

(That's not normal.)

Benjamin de Menil wrote:One question I have - are my grade nodes between the two transforms essentially working on log encoded images? Or is it that the WG/intermediate color space is is too wide gamut for my monitor, so it looks drab unless the end transform to rec709 is engaged?

Grading should indeed happen between the two conversions in the log working space. Your conversion to 709 should not be disabled of course. This is your target deliverable display which you match your monitor to.

Benjamin de Menil wrote:I also have the following question. Why is it necessary to transform to rec709 on a final node, when the timeline settings already specify an output of rec709? Am I making a mistake by doing that transform twice?

If you're using DaVinci YRGB as color management, this is a manual management mode where Resolve itself does not apply any conversions. The timeline option is only for Resolve to understand which space you decided to grade in. This matters for making sure node conversions are correctly applied. (converting nodes themselves to a different color space/gamma by rightclicking them), the color managed tools like HDR wheels and color/chroma warper and CSTs utlilize that setting if you pick Use Timeline in the dropdown. The output space you pick is there in order to let Resolve know the output space for sake of file tags on exports and the scopes.

When using Resolve Color Managed with say the HDR DWG/Intermediate preset, the conversions you'd do with CSTs are done automatically and additionally, non detected media need to have their input color space manually selected by right-clicking the media. With RCM the conversions are hidden and all grading happens in the timeline space without the ability to place nodes before or after.

There is no right or wrong approach between the two, just preference. One benefit of using RCM is that you can view your media in the source viewer with conversion already applied to it rather than viewing log.

My personal preference is manual management however since it gives the most control over what happens when and how.

Thanks, this has been very helpful.

So if I am manually converting with the colorspace transform effect, is there any need to right click the node and choose a colorspace?

Benjamin de Menil wrote:So if I am manually converting with the colorspace transform effect, is there any need to right click the node and choose a colorspace?

No, you do not need to convert any node through that mechanism. I just mentioned it for completeness sake.

The purpose of node conversions is if you want to apply some grading operations or effects in a space that is different from the grading space. The node will internally convert from the chosen project timeline space to what you pick, then all manipulations inside the node are applied and it converts back to the timeline space on it's output. This is why the timeline space needs to match the grading space you decide to use, otherwise there is a mismatch in these conversions.

A typical use case for converting a node itself is converting the gamma to Linear. I use this a lot when I want to make white balance or exposure changes which I then do with the gain wheel. This tends to give better results than using the Offset wheel when in log.

Another situation where I use a Linear node is if I need to add camera-like glow or blurs. In linear image state the luminance values are similar to the real world allowing for more physically plausible appearance when applying such effects.

The benefit of converting the node itself is simply an efficiency thing because you'd otherwise need two extra nodes as sandwich with CSTs on them to get in and back out of the space that isn't the timeline/grading space.

shebbe wrote:
A typical use case for converting a node itself is converting the gamma to Linear. I use this a lot when I want to make white balance or exposure changes which I then do with the gain wheel. This tends to give better results than using the Offset wheel when in log.

Which linear mode do you convert to for this? Sounds useful!

Benjamin de Menil wrote:Which linear mode do you convert to for this? Sounds useful!

The one that says Linear, there is only one haha.

Just remember that in order for it to be correct, the current image state where that node is placed needs to match the timeline color space.

Remember, only use gain for exposure and balance when you're in linear.

If you want to try Lens blurs you can use magic mask or depth map to create a mask for it. When using the LensBlur OFX you need to turn Highlights to 0 on the effect. It's not needed in linear light and causes artifacts.

shebbe wrote:

Benjamin de Menil wrote:Which linear mode do you convert to for this? Sounds useful!

The one that says Linear, there is only one haha.

thanks very much for this pointer! I like how the that gain wheel works now with flat gamma.