Linked / cloned / instanced layers?

[joconnell]

Hi Folks!

 

Just got my copy of photo for windows, it's all very exciting and non adobe-ish :D!

 

I work mainly in 3d so I do a lot of multi pass editing where I'll take all of the component passes of a render and rebuild them in 3d to give me the option of re-balancing, tweaking and colour correcting along the way. One thing that comes up quite often is that there's mask output from 3d to blend between different parts of the final image. These masks are often used in more than one place and should be the exact same data at all time, save for one being the inverse of the other. The idea in 3d with some things is to keep things realistic, adding more to one part of the output should automatically take away from something else. Inside of nuke it's really easy to do this as you can have a "clone" of a layer and you can also link settings. Would it be possible in Affinity to have something like a "linked layer" so that we could have one source of data as a layer (our original 3d pass) and then duplicate it but with a link back to the original layer so that any changes made to the original automatically happen in the duplicate? What we'd be able to do is have a master layer that we can paint on or colour correct as we wish and then on the duplicate layer just pop on a smart filter to invert it. Any of the changes we make to the master mask would be reflected in the other layer automatically.

 

Possible?

 

Cheers!

 

John

[Benzebub]

I second this. It would be really useful.

[Medical Officer Bones]

For inspiration how to implement this in Affinity Photo, refer to PhotoLine's virtual layers, which instance layers, layer masks and groups, groups of layers, and even adjustment layers. A master layer can be cloned, and the clone then adjusted with any of the live adjustment layers. Unfortunately independent layers cannot be linked yet, though. But I do wish Affinity Photo would have a layer stack more like PhotoLine.

Krita also supports cloned layers, but those cannot be used to clone masks. Still, the principle is identical to PhotoLine: select a layer, instance/clone the layer, and when the original source layer is edited, the instanced version updates in real-time. Even Photoshop cannot do this (although the latest version now FINALLY offers mirror painting after decades of users requesting this basic option).

Often I wish image editors with layer stacks would allow us to work with basic nodes as well, or somehow create links between layers in a more visual manner. One can dream :-)

[Aammppaa]

I've seen several threads talking about linked Masks and Adjustments over the months, and I realised that I'd never done any experimentation, but just assumed that it wasn't possible.

 

Disclaimer: I don't think this is possible in Photo, but if you have Designer then you can already do this!

 

By creating a symbol in Designer that contains a mask / adjustment layer, you can then create duplicates, which can contain different content, but retain the same mask / adjustments / both.

 

 

Takes a bit of planning, and careful monitoring of the sync button, but it does work.

 

Here is the really interesting bit… Photo can open Designer files - so you can use the same symbol (created in Designer) within Photo, even though Photo has no ability to create Symbols itself.

 

 

[kirk23]

Unfortunately, even being a step to right direction,   it doesn't work at all , at least for what original poster meant.    To be somehow useable  that "simbol" linked mask should stay in same position, keep same scale as its original.  i.e. should   have linked transforms with its original and not with simbol container,  while for some  blending operations another mask in the masking stack might still need to follow symbol container transforms. 

 

 So in fact we need not just symbols but  also transform links , like ones Photoshop has: "chain" links.      The whole approach OP asked is totally possible in Photoshop  if you use groups clipping instead of layer masks.   Neither Photoline , nor Krita  actually allow to work in that manner because of same problem: no switchable chain/transform links in between clones.   

 

But the whole Photoshop approach is so hard to manage , it turns layer stack in so huge mess of never ending smart objects stacked in groups  within a parent  group with special properties clipping another smart objects,   all of them chain linked across dozens  of other groups /smart objects, I often can't figure out anything in my own psd files.  And updating those smart objects takes forever.    Still Photoshop is totally  ready for so called "deep pixel" compositing  Nuke does.      Affinity , Photoline, Krita  doesn't.

 

And finally Affinity Photo does have cloned/ instanced layers. They call them "embedded".  Could work almost same way symbols do in Design.  You just always have to keep the original "embedded" file in the main layer stack to be accessible , then you can put clones of that embedded file withing masks of other layers and they would keep the link ( you sometimes have to scale document a bit to nudge the updating).  But not the transform link  unfortunately.  So it still doesn't works. 

 

In fact Affinity looks so close to be really useful and needs to do just a few extra steps. And it could be much better than what's in Photoshop.  I have no idea why they didn't do it while sounded promising somehow.

 

 

[Aammppaa]

@kirk23 In my demonstration I put the 2 symbols one above the other to show clearly what was happening: the mask remains linked, despite the contents being different, and everything remaining totally editable.

 

But there is no reason at all why the two symbols can't be aligned, and grouped (or placed into a parent layer), then all transforms will act equally on both.

 

So far as I can see this achieves what the OP asked for - with the possible exception that I've not addressed the "one mask should be the inverse of the other" question. However I guess that this might be done with an adjustment layer? Will have to experiment when I have the time.

[kirk23]

yeah, OP didn't asked transform links and what he asked could be done just with 'embedded" layers , a bit less convenient than with Design symbols. Still I see not that  much of a difference.

 

But its actually not enough for the purpose he described , namely editing and compositing 3d renders. making a collage of rendered ,photographed and hand painted objects.   To do so the soft should has extra option to link transforms in between layers/objects  from within different groups.   For example all objects of  a given group is following the transforms of that same group  and  a certain object within very same group is not( subtracting its pixel values from a certain mask within that group for example) and should inherit transform from totally different group. 

 

Such complicated transform links are necessary to calculate dynamic masks on the fly when you compose several 3d rendered objects  which masks  are  stacks of depth images each related to its own  objects/ group. 

   So we could set certain 3d object be always in front of another  based on it's true scene depth, then put a photo somewhere in-between , make parts out of focus also based on  true depth, have objects  intersect each other and the ground/background  plane properly without guessing  and any moment re-compose a whole collage as if it would be a 3d scene and not actually 2d.   Very same Nuke is capable for.   

 

 It's possible in Photoshop (with proper chain links setting)  and not in Affinity where we  have no means to link transforms  beyond simple groups.    And I bet the depth will soon become integral part of any photo  so first image editor that would be ready for that would definitely win the competition.

 

[Medical Officer Bones]

Correct, you list a couple of reasons that explain why I prefer to use either Fusion or the built-in compositor in Blender for 3d compositing. And those support animation: not something I would ever want to do in a layered-based image editor anyway. Layers become too convoluted and confusing when the comp becomes slightly more complex, and requires a lot of asset recycling throughout. Compare After Effects, which becomes a mess with all that pre-comping...

 

 

[kirk23]

Medical Officer Bones    Still a regular layer based image soft where you can actually draw something conveniently  with a feel of brush and canvas under your finger prints , instantly pick a layer/object of screen  etc  has it's own undisputed advantages.  Those nodes easily turn into a kind of Gordian knot and you feel as if you hold your brushes in somehow prosthetic or robotic arms.

 

  I don't need any animation too.    Just dream about a normal handy image editor having a bit more modern functionality and be not so slow ancient junk as Photoshop.  Can't tolerate Photoshop but still see no other options really.     

 

I also think the term "3d compositing" is somewhat misleading so people consider it irrelevant to their tasks.   In fact it's still regular 2d image editing ,  with just dynamic "live" masks  being calculated on the fly.  Any image editor should be capable to do so.  The math behind it is simple like 2x2 .  Games/video cards do it as "deferred"  "post effects"  on 4k screens with a speed of light.    Populating each rendered frame  i.e. compositing   gazillion pre-rendered 3d objects within a fraction of second.     So why   it's so hard to do in a regular image editor?  How many decades we still have to wait till we got something similar. It was necessary 20 years ago already.  

 

[Medical Officer Bones]

@kirk23 Totally agree, I would love to see this in an image editor other than Photoshop. Photoshop is just a very awkward and its layer stack is stuck in the nineties.

Affinity Photo, PhotoLine: both are quite close, but still missing features (which you mentioned). I've stopped waiting years ago, and now use nodal editors to do these type of jobs myself.

Still, I often ask myself why no-one integrated a "layer node" in nodal editors or an extension of nodes in a layer-based editor.