Why is Viewer Gain / Gamma post ViewLUT?

Only just noticed this as I don't work all that often in Fusion but is there any logical reason that the viewer's gain / gamma controls are post ViewLUT? This makes little sense to me from a compositing standpoint and I believe no other compositing software operates like this, even After Effects.

When working in linear you want to be able to gain down/up the viewer to see if you still have info in the highlights/shadows while viewing under a DRT. It feels like this operation order was never changed because simple linear to gamma conversions without tone mapping were expected in which case gain would still work. But such practices are long gone in modern day compositing.

I would hope that BMD can hard switch this order of operations or give an option in the prefs to flip it.

Agreed,

You cannot adjust exposure post viewer transforms. The whole point is to adjust stops linearly.

Please reorder the operations so that the viewer lut is the LAST operation.

I works this way as it always has. And its super flexible and useful the way it set up. So many options for any possible situation you would want including A/B split for each of the two viewers, including an even quad view for everything you would ever want to see while working while previewing final results. I don't see any problems with this.

sshot-3921.jpg (271.21 KiB) Viewed 762 times

sshot-3922a.jpg (298.85 KiB) Viewed 762 times

sshot-3922b.jpg (202.1 KiB) Viewed 762 times

And its super flexible and useful the way it set up.

Perhaps you're missing my point. I understand that you can have multiple viewers and A/B within a viewer with each a different ViewLUT but my issue is solely the order of operation of the Gamma/Gain adjustment you can make on the viewer. It's in the wrong place to be practical for true scene referred linear workflows. Converting a Rec.709 normalized image to linear and back is not really scene linear. The current behavior is undesired for working with linear CG renders and/or scene referred camera data like camera log or raw.

Agreed, the order of operations is wrong and very misleading.

And its super flexible and useful the way it set up.

Perhaps you're missing my point. I understand that you can have multiple viewers and A/B within a viewer with each a different ViewLUT but my issue is solely the order of operation of the Gamma/Gain adjustment you can make on the viewer. It's in the wrong place to be practical for true scene referred linear workflows. Converting a Rec.709 normalized image to linear and back is not really scene linear. The current behavior is undesired for working with linear CG renders and/or scene referred camera data like camera log or raw.

I must be missing something indeed. I don't see the problem you are referring to. Its just a matter of setting up your composition. Perhaps you can post an example of what you want, what you think is this wrong order in Fusion and I can take a look. Either I'm wrong or you are wrong. We can find out together.

You can use log or rec709 or whatever you wish.

Its just a matter of setting up your composition.

Viewer operations are not part of the comp by design, so it is not a matter of setting up the comp graph. Adding nodes explicitly to the graph to do what viewer is supposed to be doing is just wrong. And it is wrong because viewer controls are meant for display purposes, not for modification of the actual data in the stream. Adding nodes does just the latter, which is not desirable behavior in any way. Viewer gain/gamma, LUTs, channel toggles, mask overlays etc can all be done with nodes, sure, but the sole reason why they are in viewer is to allow not throwing nodes left and right.

Thanks for the added explanation Hendrik.
Indeed viewer processes should not be part of the actual node tree otherwise you keep endlessly fiddling with them.

Perhaps you can post an example of what you want, what you think is this wrong order in Fusion and I can take a look.

Here's an example to illustrate the issue. The source is a linear exr.
With the current order you are unable to reveal the information in the lights in the background. If you do this in Nuke, After Effects, Natron, Autograph etc. the order/behavior is the one on the right.

I find it difficult to come up with an argument against switching the order, so hopefully if BMD reads this they can implement it in a future update. Would imagine it's easy to fix..

viewLutOrder.jpg (792.94 KiB) Viewed 361 times

Thanks for the added explanation Hendrik.
Indeed viewer processes should not be part of the actual node tree otherwise you keep endlessly fiddling with them.

Perhaps you can post an example of what you want, what you think is this wrong order in Fusion and I can take a look.

Here's an example to illustrate the issue. The source is a linear exr.
With the current order you are unable to reveal the information in the lights in the background. If you do this in Nuke, After Effects, Natron, Autograph etc. the order/behavior is the one on the right.

I find it difficult to come up with an argument against switching the order, so hopefully if BMD reads this they can implement it in a future update. Would imagine it's easy to fix..

viewLutOrder.jpg

Hmm. I still don't understand what the problem is with Fusion, since for me this is no issue because I work differently than your set up and doing what you want to do is easy. But I can see how the way you work it would be a problem for you. I guess like so many on this forum, there are people who want the application to work as they work. If Black magic implemented whatever you think is correct, I would not mind it, as long as there are options to work as it works now. Because otherwise I would find it a problem. But to each their own I guess. There is nothing to fix as far as I can see, only to accommodate.

Hmm. I still don't understand what the problem is with Fusion, since for me this is no issue because I work differently than your set up and doing what you want to do is easy.

So how exactly do you switch the order of viewer operations without adding nodes to graph that modify the data?

There is no "setup" for the viewer, its ops are hardcoded. Juggling nodes isn't part of the problem nor solution, it is a totally different area.

Hmm. I still don't understand what the problem is with Fusion, since for me this is no issue because I work differently than your set up and doing what you want to do is easy.

So how exactly do you switch the order of viewer operations without adding nodes to graph that modify the data?

There is no "setup" for the viewer, its ops are hardcoded. Juggling nodes isn't part of the problem nor solution, it is a totally different area.

Its simple really. You use one Gamma or CST (Color Space Transform) Node at the beginning and one at the end of the node tree. Assuming you are using manual approach, and not color managed workflow that is also available in Fusion. But if you want total control and managed it yourself, which is what I personally prefer, you use it to transform everything to linear space on input and out of linear space to whatever you like at the output. That way Fusion works as intended in linear space. If your footage is already linear than you do not need to convert it on input.

That takes care of the way nodes work. Linear. I set up one CST node for most common footage I work with and set it to default so all I have to do is click on the shortcut for it on the toolbar and I'm ready to go on the input. For output I just copy the same node and click the swap button and I'm back where I need to be. Simple and quick once you set it up.

But this may darken the image in the viewer(s) or if its log lighten it etc. So we need viewer luts to deal with what we are looking at to be correct, while fusion gets its linear footage that it wants, since most tools in Fusion are designed to expect linear input and work the best that way.

For the viewer luts, depending on how you work its also one click button and or list of common luts for most types of footage that you can set up. so its again, easy. You can use view luts or not at any point. The other one is buffer lut for 3D work, but its essentially the same process. No problem.

When working with Gamma / Gain it simply exposes or hides a simple pair of Gain and Gamma sliders that let you adjust the viewed image without altering it. Especially useful for “gamma slamming” a composite to see how well it holds up with a variety of gamma settings. Defaults to no change. As one might expect.

I have never had a problem with it. If you do, you probably should change your workflow.

What Fusion allows is very complex and user controlled workflow that can fit any need of any input or output you might have. Linear, Log, Rec709, HDR, SDR, or combination of multiple inputs and outputs for technical evaluation, conversion needs or for creative purposes. Its super flexible. Nothing needs fixing.

As for order of operations, here is from the manual.

sshot-3952.jpg (108.75 KiB) Viewed 343 times

P.S. When working in Fusion studio and when Working in Resolve things are bit different, because in resolve you can also do bunch of stuff in color or edit page so that comes into play. But process is not a problem if you know what you are doing.

I have never had a problem with it. If you do, you probably should change your workflow.

It seems you don't care about it's current behavior and thus deem it as not problematic. To say that I should change my workflow is odd. Maybe tell that to the entire vfx industry not working Fusion and ask if they agree.

I set up one CST node for most common footage I work with and set it to default so all I have to do is click on the shortcut for it on the toolbar and I'm ready to go on the input. For output I just copy the same node and click the swap button and I'm back where I need to be. Simple and quick once you set it up.

This is not how you want to manage. The whole point of a viewLUT is that the output stage of the entire node tree remains in the working space (which we make scene linear for best results). Otherwise you cannot see the image in it's output state while viewing any of the nodes prior to your output conversion if you purpose the node tree itself for that. By your logic they may as well undo the v19 update to the viewerLUT that added OCIODisplay and CST as options.. which annoyingly still misses the ACESTransform and OCIOLook.

Changing gain in the viewer should act as exposure on linear data so needs to be before the viewLUT is applied and not after. The order of ops in the manual does not mention the viewer gain/gamma sliders but they are after the LUTs. It should be directly after the Image Output so it's applied on the data before viewer adjustments are made.

What Fusion allows is very complex and user controlled workflow that can fit any need of any input or output you might have. Linear, Log, Rec709, HDR, SDR, or combination of multiple inputs and outputs for technical evaluation, conversion needs or for creative purposes. Its super flexible. Nothing needs fixing.

The type of workflow deployed inside Fusion has little to do with viewer gamma/gain.

I guess like so many on this forum, there are people who want the application to work as they work.

Sure, but there are things we could categorize as industry standard, or naturally expected because the majority needs it. Like proper scalable and adjustable UI which some people on this forum seem to defend against under the argument that it works for them and make similar claims that people should just change their workflow or buy other monitors.

This is not how you want to manage. The whole point of a viewLUT is that the output stage of the entire node tree remains in the working space (which we make scene linear for best results).

That's incorrect.

Otherwise you cannot see the image in it's output state while viewing any of the nodes prior to your output conversion if you purpose the node tree itself for that. By your logic they may as well undo the v19 update to the viewerLUT that added OCIODisplay and CST as options.. which annoyingly still misses the ACESTransform and OCIOLook.

You don't understand what I explained so your comment reflects that.

Changing gain in the viewer should act as exposure on linear data so needs to be before the viewLUT is applied and not after. The order of ops in the manual does not mention the viewer gain/gamma sliders but they are after the LUTs. It should be directly after the Image Output so it's applied on the data before viewer adjustments are made.

You don't understand the purpose of the tool. Nor how to use it. So no wonder your are frustrated. Fusion and me are just fine.

The type of workflow deployed inside Fusion has little to do with viewer gamma/gain.

Its a useful tool to preview where the problems are. No more and no less. It does not change the image in any way. Which is the whole point of the tool.

Sure, but there are things we could categorize as industry standard, or naturally expected because the majority needs it. Like proper scalable and adjustable UI which some people on this forum seem to defend against under the argument that it works for them and make similar claims that people should just change their workflow or buy other monitors.

I guess there are always people who have a problem for every solution. I can't help you with that attitude. Sorry.