V-Log to linear and BM Film Log to linear

Moderator: SecondMan

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#16

Post by Wizor » Sun Jul 14, 2019 10:24 am

So in the first example where i added srgb gamma, if i want output rec 709 how can i do this? i dont see rec709 output in gamut node(i want this comp for delivery i think? i dont want do more operations after that compositing/grading in Fusion)
What output format i could use to save compositing for future editing/grading?(theoretical question)

Same as in OCIO, in gamut transform i dont see BMD option.

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#17

Post by gez » Sun Jul 14, 2019 11:47 am

Sorry, I replied to the last comment but missed the previous one with the captures.
Both are close but not there.

In the first one, you addessed the transfer curves properly, but you missed the primaries. Remember that a colorspace is comprised by the transfer, the primaries and the white point (check my first reply).
You need to add a gamut node to each camera input in order to transform from each camera's primaries to rec.709.

In the second one (ACES), the chain of transforms is ok, but as you already noticed the IDTs are missing and you used ACEScc in a couple of nodes.
Remember that the idea is to take all the inputs to the same reference colorspace, and in the case of compositing the most appropriate is ACEScg if you're using the ACES workflow.
So, for cameras:
- IDT of Panasonic camera to ACEScg (available in ACES 1.0.3)
- IDT of BMD camera to ACEScg (missing from ACES 1.0.3, so you need to add that transform via an alternate OCIO config or using the method I described in my previous reply)
- Transform fron Blender's linear rec.709 to ACEScg (as it's not a device, you won't find a device. I use in this case the utility - linear rec.709 option as input)
After that, all your compositing will be done in linear ACEScg colorspace, and the result will be converted to the needed output (you chose rec709 in your screenshot, which is fine if you want to output to HD video).

Again, one thing you have to ask yourself here is why you want to use ACES. Is if future-proof? Is it multiple outputs with wide-gamut in mind?
If you can't answer that it's possible you don't need ACES and sticking to rec.709 instead is a better idea.

User avatar
Midgardsormr
Fusionator
Posts: 1061
Joined: Wed Nov 26, 2014 8:04 pm
Answers: 1
Location: Los Angeles, CA, USA
Been thanked: 63 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#18

Post by Midgardsormr » Sun Jul 14, 2019 12:16 pm

They used a different name for 709 in the Gamut node. I don't have Fusion in front of me, but I think it's something like ITU-R BT.709. I never can keep it clear in my head if the output is supposed to use the display or scene variant, though.

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#19

Post by Wizor » Sun Jul 14, 2019 12:25 pm

Can u make screen from Fusion how this workflow should looks like? Im really confused right now becouse i dont know how to apply this inside program.
I thought after remove log curve i can start compositing and just at the end i need convert it to output color space to propely display it on monitor...
Which gamut(source/output space) i need to put after Cineon node to each camera?
(i dont see BMD in gamut tool to transfer it in to ACES)

@Midgardsormr thanks. you remembered it correctly, next question is (display) or (scene)?

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#20

Post by gez » Sun Jul 14, 2019 1:36 pm

@Midgardsormr I'm always in doubt about those two, so I just asked Troy.
Rec.709 (scene) is only intended to produce a display linear rec.709 from broadcast cameras or cameras that produce true rec.709 recording. It uses a 1.96 transfer.
Rec.709 (display) is intended for display, as the name states, so if you're going from scene linear rec.709 to HD delivery, you want that one in the output. gamma 2.4

That being said, most of the prosumer cameras, DSLR and cheap cameras everyone have at home, record HD but with the display transfer baked (2.4) and some contrast curve, other sweeteners and tonemapping, so the display-linear produced by the transform won't have radiometrically linear ratios from the scene, hence useless for VFX and compositing.
The same applies if you're ingesting some video from the web, which is most likely encoded for display and won't cut it for compositing because all of the baked aesthetic stuff that can't be reverted to anything close to the scene originally shot.

TL;DR: If you have footage from a rec.709 proper camera, the "scene" variant is the right one to ingest and linearize your footage as display-referred rec.709 in your compositing pipe and the "display" variant is only intended for producing a display-ready output.

@Wizor if you won't use ACES (and I insist that you shouldn't unless you have a good reason for it), just follow these guidelines (for rec.709):

Your Blender renders (from EXRs) are already linear rec.709, you don't need to use any colorspace transform

Your camera footage has its own primaries and transfer, you have to deal with them this way:
- With a CineonLog node, go from Log to Lin choosing your camera's log.
- Next, with a Gamut node, go from you camera's primaries to rec.709 (it won't matter which one you choose, scene o display, as you won't mark any of the gamma nodes as footage is already linear).

All your compositing will be done in linear rec.709 colorspace.

When you're done, add a gamut node to the tail, leave the source colourspace unchanged, and choose Rec.709 (display) as Output space.

Done, that can go to your saver to produce a display-ready HD video.

User avatar
Midgardsormr
Fusionator
Posts: 1061
Joined: Wed Nov 26, 2014 8:04 pm
Answers: 1
Location: Los Angeles, CA, USA
Been thanked: 63 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#21

Post by Midgardsormr » Sun Jul 14, 2019 3:57 pm

gez wrote:
Sun Jul 14, 2019 1:36 pm
so the display-linear produced by the transform won't have radiometrically linear ratios from the scene, hence useless for VFX and compositing.
I think that's too strong a statement. VFX doesn't typically require perfect physical accuracy. If it looks right, it is right. Everything we do is a kludge, anyway, so close enough is often good enough.

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#22

Post by gez » Sun Jul 14, 2019 4:21 pm

Midgardsormr wrote:
Sun Jul 14, 2019 3:57 pm
gez wrote:
Sun Jul 14, 2019 1:36 pm
so the display-linear produced by the transform won't have radiometrically linear ratios from the scene, hence useless for VFX and compositing.
I think that's too strong a statement. VFX doesn't typically require perfect physical accuracy. If it looks right, it is right. Everything we do is a kludge, anyway, so close enough is often good enough.
Yes, you're right. I guess that "useless for VFX and compositing" was going a bit too far. You can use them of course, and everyone does it, myself included (I'm a graphic designer, so 99.9% of the source material I use is display-referred, and it ends up in composites :-)

However it's important to note that every display referred plate you put in your composites is going to bring some extra work to make it "good enough" that usually you don't have to do when you deal with scene-referred materials.
Dynamic range is the first thing to pop up: You found that beautiful JPG you want to use as background plate, and it turns out that you can't dial the exposure down a bit because the sun turns gray.
You put a lens blur effect and it looks horrible, not a lens blur at all, and so on.
Using a display-referred image in a scene-referred pipe is like using cardboard billboards with printed textures. Sure, in some situations it can cheat the viewer and it does look like you have a photo with Obama, but when the camera moves a little or lighting changes the effect fallls apart.

So yes, maybe saying it's useless was an exaggeration, but usually people cant' see the issues (and the extra work) that using a display-referred asset will carry with it.

On the other hand, a properly managed scene-referred workflow brings more simplicity and many things you struggled with back in your photoshop days tend to go away, with a lot of goodies for free. The lens blur example I used above is one of them, once you see it, you can't unsee it :-D

Fair? :-D
Last edited by gez on Sun Jul 14, 2019 8:55 pm, edited 1 time in total.

User avatar
Midgardsormr
Fusionator
Posts: 1061
Joined: Wed Nov 26, 2014 8:04 pm
Answers: 1
Location: Los Angeles, CA, USA
Been thanked: 63 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#23

Post by Midgardsormr » Sun Jul 14, 2019 6:45 pm

gez wrote:
Sun Jul 14, 2019 4:21 pm
Fair
Quite!

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#24

Post by Wizor » Mon Jul 15, 2019 4:18 am

Okay, collecting all information from your posts i updated these 2 graphs for Linear and ACES workflow.
Tell me if everything is okay now, I have marked places in which i still struggle.
LinearWorkflow_Update.png
ACES_Workflow_Update.png


Question outside the topic: when i should key greenscreen? Before or after converting footage to linear/aces?
You do not have the required permissions to view the files attached to this post.

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#25

Post by gez » Mon Jul 15, 2019 6:31 am

Wizor wrote:
Mon Jul 15, 2019 4:18 am
Okay, collecting all information from your posts i updated these 2 graphs for Linear and ACES workflow.
You're getting closer :-D There are a few things to tackle yet.

And btw (and sorry for pestering with terminology): Both workflows are linear workflows. In the past when people started to talk about "linear workflow" it usually meant linear rec.709 because many apps offered nothing but tat, but in a broader sense a "linear workflow" means that you're working with rediometrically linear images. Those could be encoded with any primaries.
Now that we have OCIO and we can switch references and colorspaces easily, saying just "linear workflow" is probably incomplete and doesn't communicate accurately, so in this case it's better if you also mention the primaries: Linear rec.709
Wizor wrote:
Mon Jul 15, 2019 4:18 am
Tell me if everything is okay now, I have marked places in which i still struggle.
Ok, first screenshot. Everything is in place, but your gamut nodes need some tweaks:
- Deselect all the "remove gamma" checkboxes. Your transfer was removed with the log to lin node already, so it's no longer necessary.
- Although metadata from the loaders should take care of this, it's better if you manually assign the source colourspace in the "input" gamut nodes. So you have to set BMD and Panasonic colorspaces in the Source Space. It will work for Panasonic, but there's no BMD gamut :-(
- You shouldn't mark "add gamma" in those nodes. Remember, you want to keep it linear for compositing, that's why you removed the log in the previous step.
- In the output gamut node uncheck the "remove gamma" checkbox. Again your stuff is linear, there's no curve to remove. In this case it's correct to add the gamma to the output, because in that stage you want to go from linear to gamma-corrected.
- As you guessed, a gamut viewLUT will help during the compositing to display the linear data properly. Uncheck remove gamma (already linear), check add gamma (you want it non-linear for display).

Second Screenshot (ACES). Everything looks ok, except the missing BMD IDT/Gamut :-(
It looks like BMD wants you to use Resolve's CM to take care of that.
https://forum.blackmagicdesign.com/view ... =2&t=87576
https://forum.blackmagicdesign.com/view ... 21&t=78109
It's a mess. I'll investigate what's the right approach, but I'm sure that somebody more experienced than me with BMD hardware knows what to do. Meanwhile a quick and dirty workaround could be to re-encode the BMD footage with Resolve and ingest it in a different colorspace. Not ideal but it should work.
Wizor wrote:
Mon Jul 15, 2019 4:18 am
Question outside the topic: when i should key greenscreen? Before or after converting footage to linear/aces?
There are different approaches. Some people pulls a matte from the original shot and others convert to reference first.
Personally I think it's more correct to do it with the original plate to avoid crosstalk. After the colorspace transform some pixels in the foreground/greenscreen edges migh be a wee bit distorted (color-wise) and make more difficult to pull clean edges.This is specially visible in hair.
If you do it BEFORE the colorspace transform remember to check pre-divide/post-multiply alpha in the colorspace transform operation.

A final word of advice: It's always a good idea to keep color grading after compositing, as a finishing step. Also, grading on log data is usually the way to go.
It doesn't mean that you can't do some color correction in linear space, but it will help keeping your project tight and easier to read and edit if you leave the grading separated from compositing. This is specially important when you have to hand your work to a colorist.
I know I'm probably adding an extra layer of complexity to something you're just learning, so just keep it as a note for now :)

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#26

Post by Wizor » Mon Jul 15, 2019 7:52 am

I thought "remove gamma" checkbox does nothing if color space is not assigned(when i removed curve log to lin i see color value changes but with this checkbox nothing really happen).

"Although metadata from the loaders should take care of this" so i dont need define V-log and BMD(BMD not even on list) gamuts becouse when its on "no change" its take source file color space yup? It means first graph with primaries: linear rec.709 :D is just fine i if uncheck "remove gamma" in all gamut nodes ^^

Also if gamut node source space "no change" take care about color space from loaders, isnt it fine to do in ACES Workflow:
do gamut node after loader looking like: source:no change to output space: ACES(no checkboxes for remove and add gamma)
and after that add OCIO: ACES to ACEScg?(i dont have an Fusion open right now but i dont know if alone "ACES" IDT/Gamut even exist)

Hmm isnt it wierd BMD software like a Fusion havent a BMD gamut/ocio opiton?^^


And i would like to add huge thanks for you, its difficult subject and sometime its hard to find information and explanations about it, someone should make good tutorial about handling color spaces in Fusion to explain this.

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#27

Post by gez » Mon Jul 15, 2019 1:33 pm

Wizor wrote:
Mon Jul 15, 2019 7:52 am
I thought "remove gamma" checkbox does nothing if color space is not assigned(when i removed curve log to lin i see color value changes but with this checkbox nothing really happen).
It depends, if your input still has a colorspace metadata it could remove the transfer again.
It shouldn't and it probably doesn't happen, I just don't trust automatic stuff with metadata in an environment like a compositor, where you can arbitrariy modify the data.
Wizor wrote:
Mon Jul 15, 2019 7:52 am
"Although metadata from the loaders should take care of this" so i dont need define V-log and BMD(BMD not even on list) gamuts becouse when its on "no change" its take source file color space yup? It means first graph with primaries: linear rec.709 :D is just fine i if uncheck "remove gamma" in all gamut nodes ^^
Hmm, let me clarify:
When a loader gets metadata from the file, the default "auto" setting in theory does the right thing, the node passes the metadata, and the nodes should be aware of the colorspace and other relevant image details.
However, some file formats don't include the type of metadata needed.
Conversely, the OCIO nodes just ignore metadata.
It's a bit of a mess, so you have to be keep an eye on it. It's up to you what to do, but personally I don't trust in automatic stuff happening behind the scenes, so I prefer to set the source and destination colorspaces.
The only exception is when I know that all my composite is linear rec.709 already and I want to prepare it for output, where ignoring the input and adding the rec.709 gamma wih a gamut node is fine.
Wizor wrote:
Mon Jul 15, 2019 7:52 am
Also if gamut node source space "no change" take care about color space from loaders, isnt it fine to do in ACES Workflow:
do gamut node after loader looking like: source:no change to output space: ACES(no checkboxes for remove and add gamma)
and after that add OCIO: ACES to ACEScg?(i dont have an Fusion open right now but i dont know if alone "ACES" IDT/Gamut even exist)
ACES workflow and Fusion's color workflow are different. In ACES you leave all the transforms to ACES via OCIO.
When you convert from, say Panasonic VLog to ACEScg, the IDT deals with the primaries AND the transfer in the same step, resulting in a linear ACEScg buffer.
From that point it's all ACES until you define an Output.
I suggested to use a hybrid hack assuming that there was a colorspace transform from BMD gamut to ACES, as a preliminary step before ACEScg via OCIO.
But it's not the case, so forget it :-D
Wizor wrote:
Mon Jul 15, 2019 7:52 am
Hmm isnt it wierd BMD software like a Fusion havent a BMD gamut/ocio option?^^
It's quite inconvenient. Weird? I don't know. I think it reflects their intention to push users into leaving Resolve deal with the CM.
I don't know who this works/is going to work in Fusion 16 Studio, but it's quite inconvenient that the software can't deal with the same brand camera footage.
Wizor wrote:
Mon Jul 15, 2019 7:52 am
And i would like to add huge thanks for you, its difficult subject and sometime its hard to find information and explanations about it, someone should make good tutorial about handling color spaces in Fusion to explain this.
You're welcome. I've been thinking about writing a tutorial or even making a video and other have asked it, but I haven't found the time to do it.
It's on the map though. I just want to do ir properly and that takes work.

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#28

Post by Wizor » Wed Jul 17, 2019 3:46 am

Ok i have another problems here but with grading workflow. After converting these footages to linear both are very overexposed(values much above 1, BMD even more than V-log).
I dont know how should look correct workflow here for grading, should i clamp these values to 0-1 and start working from here?
Second problem is white balance, i have footages with color charts but the colors on them are different, tried white balance node to match it but i still have diffrences in color consistency (both footages records same subject just one is wide and one close up).

User avatar
gez
Fusioneer
Posts: 140
Joined: Mon Jul 16, 2018 6:21 pm
Location: Argentina
Been thanked: 15 times
Contact:

Re: V-Log to linear and BM Film Log to linear

#29

Post by gez » Wed Jul 17, 2019 7:16 pm

Wizor wrote:
Wed Jul 17, 2019 3:46 am
Ok i have another problems here but with grading workflow. After converting these footages to linear both are very overexposed(values much above 1, BMD even more than V-log).
I dont know how should look correct workflow here for grading, should i clamp these values to 0-1 and start working from here?
Second problem is white balance, i have footages with color charts but the colors on them are different, tried white balance node to match it but i still have diffrences in color consistency (both footages records same subject just one is wide and one close up).
At this point that's a bit off-topic and you should probably start a new thread with your grading questions and some screenshots showing your setup and some sample images to visualize the problem better.
As an advance I'll only say that clamping is not the best idea, going to a log intermediate for grading should be better. But that's for another thread. :D

BTW, how did you manage to convert your BMD footage to linear without the transforms?
Last edited by gez on Thu Jul 18, 2019 7:14 am, edited 1 time in total.

Wizor
Posts: 11
Joined: Mon Jun 10, 2019 11:11 am

Re: V-Log to linear and BM Film Log to linear

#30

Post by Wizor » Wed Jul 17, 2019 11:58 pm

So i should grade them before converting to linear? Removing curves to linear will blow up grade of these footages i guess.

For BMD Film log i i did first cineaon node (BMD Film log to lin, like on my screeshots) and after i put gamut with source space:no change to output 709 , but now i noticed its does not make any difference ;c