P: Add Layers to Lightroom

[bcw99]

I've seen a plugin that adds layers to LR which would save a lot of to-ing and fro-ing to Photoshop. The plugin is actually stand-alon, but also integrates with LR to some extent. It allows many of the layer options found in Photoshop. Not tried it but seems like a cracking idea! 🙂

Making LR more of an editor could make Photoshop redundant for pure photographic work

[dorin_nicolaescu]

Forked to re-merge into a more appropriate thread. Please reference the new topic here: Lightroom: Preserve PSD Layers in LR to allow back and forth editing

[hillrg]

An excellent summary TK. I frequently wonder why "pundits" push this pixel pushing vs parametric editing idea. Andrew must have been having a bad day when he "took umbrage".

[TheDigitalDog]

>>I frequently wonder why "pundits" push this pixel pushing vs parametric editing idea.

Because we actually understand the differences between the two processes and the results each produces. We actually understand the role of differing applications as tools and use them on the data they were designed to process.

[FeedbackCommunityMember]

Andrew, do you have a technical argument as well?
How do you counteract the technical arguments I've made?

The differences between "the two processes" are current de-facto differences that are historically rooted and are only weakly justified by technical arguments.

Are you rejecting adjustment layers in PS because they are not "pushing pixels", but are an example for parametric editing? Why are adjustment layers in PS not an inadequate mixing of "two processes"? Have adjustment layer been added to PS by someone who does not "understand the difference between the processes and the results each produces"?

[TheDigitalDog]

I already told you nearly a year ago that layers are NOT non destructive. Well if you decide you want to do anything with the image outside of Photoshop or print it, or convert it to another color space the layers (edits) ARE applied and boom, you've got a destructive edit.

Parametric instructions are just that. They tell a raw processor HOW to CREATE RGB pixels. That is non destructive. You're not editing pixels, you're creating them. But in the year since we last talked, you don't see to have moved past the differences which are rather significant.

[FeedbackCommunityMember]

Andrew, my sincere apologies for forgetting about the long debate we had. I would not have asked you any questions, had I been aware of the debate.

However, my points stand and you are mistaken on the nature of adjustment layers. You claim they are "NOT non destructive", but clearly they are non-destructive. The reason is that you can revisit the parameter settings of each adjustment layer at any time and any changes to the parameters will be reflected in a new rendering.

Imagine a layer stack of five adjustment layers, each controlling some different aspect (such as hue, levels, posterisation, etc). You can change any of the layer parameters at any time and the overall effect is exactly like the LR experience. You can change sharpness, levels, colour, etc. at any time in any order.

The adjustment layers of PS "create" pixels just like a step in the LR image pipeline "creates" pixels.

This is what parametric editing is about; that you can always revisit your decisions and change them without losing any edits you have done after you made your initial parameter choices.

The non-destructive nature of adjustment layers is their raison d'être. Before adjustment layers, you had to bake in any adjustment (such as a curves adjustment, for instance) into your image and you couldn't go back, change the parameters (e.g., the curve) and enjoy the different result (without losing, for instance, hue adjustments you made subsequently to the curves adjustment).

The fact that the effect of adjustment layers has to be applied once you export a file ("...do anything with the image outside of Photoshop...") is completely immaterial. It exactly corresponds to the "export" step in LR. Are you saying, just because LR can export images and therefore "boom" has to become "destructive? makes it lose its non-destructive status?

P.S.: I should perhaps add that PS adjustment layers are only non-destructive as long you do not combine them with regular destructive layers. The latter will of course be able to override the pixels the adjustment layers "created" (to use your terminology) or create copies of them that won't be updated anymore upon further changes to the adjustment layer. But if you stick to only using adjustment layers, you can replicate a non-destructive LR experience in PS. Again, has this been made possible by a person/team who does not truly understand the difference between the -- according to you so fundamentally different -- "processes"?

[TheDigitalDog]

>You claim they are "NOT non destructive", but clearly they are non-destructive.

Clearly? Think about the process and I think you'll see it's quite clear layers are destructive. What do you think happens when you print the document, the layers somehow disappear? That the edits are NOT applied? Of course they are. And there IS data loss. No way around that (and as I said a year ago, moot in high bit but data loss none the less).

The ONLY time adjustment layers are NOT destructive is when you view the image in Photoshop. Flatten, what happens? The edits are applied. You can't edit pixels and change their values and not lose data to rounding errors. The second you do anything like print, flatten, save as a different color space etc, the edits IN the layers ARE applied to the underlying data. There's no other way around this fact you find so unclear!

>>The adjustment layers of PS "create" pixels just like a step in the LR image pipeline "creates" pixels.

No, they don't. You seem to misunderstand not only the processing of the data in these two applications, you seem to not realize that LR cannot alter pixels, it has to render them from sets of instructions. Rendering is the process that takes original data, instructions and creates a new set of pixels from both. With a raw file, that's non destructive. You are (for at least the 3rd time), creating pixels. IF you take a rendered image (TIFF) run it though LR.'s engine, the original is untouched but the rendered iteration is not a non destructive action if you take this edit and compare it to the original data. But we want to alter pixels so you either do so an end up with some data loss or you leave the pixels alone. Adjustment layers alter the pixel value (or you'd see nothing happening).

>This is what parametric editing is about; that you can always revisit your decisions and change them without losing any edits you have done after you made your initial parameter choices.

Yes, ONE part of the process. But look what you wrote and think about it with what I've written: you can always revisit your decisions yes, and since you must render a new iteration, it's original data + instructions equals new iteration from a rendered image and a new creation from raw.

>It exactly corresponds to the "export" step in LR.

No it doesn't. Not even close. Export is a rendering where again, new pixels are created from two possible data sources. In Photoshop, you're directly altering pixel values. Whether you do it without a layer or with a layer, the net results at any time expect just viewing the image in Photoshop is the same.

Alter existing pixels= data loss. Creating pixels is completely different. In high bit, the data loss isn't visible but the data loss is there and one can view the differences between the two easily using the Subtract command with both docs.

[FeedbackCommunityMember]

You write "In Photoshop, you're directly altering pixel values., but this is wrong for adjustment layers.

Andrew, do yourself a favour and start PS, and create a background image with differently coloured brush strokes. Then stack a B&W adjustment layer on top of it. Now stack a curves layer on top and make it create a very high contrast. By manipulating the B&W adjustment layer alone, you can now make brush strokes appear and disappear, because the curves adjustment layer will respond to the new pixels that the B&W has "created". Nothing ever gets destroyed. Your original brush strokes are always untouched. There is only transformation (as in LR), never destruction.

With adjustment layers, there won't ever be any data loss due to direct pixel manipulation as you described. Why is that? Because PS (just like LR) "creates" pixels (as you call it) when producing the result of an adjustment layer. If it didn't how could a change at the bottom most layer ripple up through further adjustment layers? If the curves adjustment layer in the example I mentioned above deemed a pixel value to become pure black in one instance, how can it transform it to white, even though you only changed the mapping from colours to gray values in the adjustment level below? The reason is that the curves adjustment layer works on pixels that the B&W adjustment layer "created". You could say "re-created" based on new parameters. This is no different to what happens in the image pipeline of LR.

We can avoid references to the image pipeline altogether, though. Just explain to me how the user experience you can have in PS with adjustment layers only is any different to what the user experience in LR is. Only when you take an image outside PS (or make a destructive step like flattening layers), you commit to certain adjustment layer values, just like when you commit to certain edit parameters when you export an image from LR.

Surely flattening layers (or other destructive operations) makes the editing destructive but that is completely besides the point. I never claimed that all of PS is non-destructive (it isn't by any stretch of imagination). You are insinuating that I don't understand that "print, flatten, save as a different color space etc" implies an application of the layers to the underlying data. Of course, I understand that and at this point of course the non-destructive train that adjustment layers provide stops. But the very same is true for LR. Only as long you stay within LR, the non-destructive train is running. Export an image, and there is destructive pixel baking just like with "save as" (or similar) in PS. Obviously, the "destructive pixel baking" does not refer to the raw original. Just like the original brush strokes you made in you PS background layer will never be altered. Think about this Andrew: You'll be saving transformed versions of the original brush stroke pixels when you do a "Save for Web" for example. But PS will never destroy your original brush stroke pixels and any stacked adjustment layers will always be fluid in how they respond to the untouched brush stroke pixels. "Destructive pixel baking" as I used it above only refers to the exported image since you cannot take it and revisit any editing choices you made. Same for PS and LR.

[dorin_nicolaescu]

Adjustment layers? I thought this thread is about real, pixel layers...

Aren't Lightroom's local corrections the same as Photoshop's adjustment layers? Except they are not called so and don't look like that.

[FeedbackCommunityMember]

Dorin, regarding your first question: Yes, the thread is about layers, but as some (like Andrew) are arguing that PS and LR are fundamentally different (and layers hence have no place in LR), I used adjustment layers as a vehicle to illustrate what the real essence of non-destructive editing is (and that hence things that are often considered to be reserved to "pixel pushing" applications, like retouching, could very well be a part of LR without breaking its non-destructive paradigm).

Regarding your second question: Many LR adjustments (including global ones) are like PS adjustment layers, sometimes combined with implicit layer masks (in particular when local corrections are concerned). Note that there is a one-to-one correspondence between the "layer on/off"-eye symbol for a PS adjustment layer and the "on/off" switch for an LR panel. The similarities are really in one's face, but Andrew appears to remain unconvinced.

[TheDigitalDog]

>>Andrew, do yourself a favour and start PS, and create a background image with differently coloured brush strokes. Then stack a B&W adjustment layer on top of it. Now stack a curves layer on top and make it create a very high contrast. By manipulating the B&W adjustment layer alone, you can now make brush strokes appear and disappear, because the curves adjustment layer will respond to the new pixels that the B&W has "created". Nothing ever gets destroyed.

OK, let's back up since we may have a semantics issue here.

Let's take your suggestion above. You say nothing has been destroyed. So let's make this simple. Take your above Photoshop layered doc. What happens in terms of destructive editing to pixels if I flatten the doc? Wait, you answered this:

>Surely flattening layers (or other destructive operations) makes the editing destructive

OK, so just what do you suppose happens if:

I don't flatten but I print the document in Photoshop? The underlying pixels do not get affected? Of course they do. What if I convert to another color space? Either the pixels and layers are all combined to produce the new appearance or none of the numbers are affected in which case, explain how the image now appears different.

You have a single pixel with a value of R35/G55/B65 OK. Now I pop an Adjustment layer on top such that the values will be R37/G55/B65. Only ONE value has changed. Your take on the PS processing is, the original values are sent to the printer or the NEW values? If the NEW values are sent, it's a destructive edit due to the rounding errors.

Maybe the version of Photoshop the people on Star Trek use can alter color and tone appearance of digital images without altering any values but that isn't how image processing works in 2013.

If I apply an edit in a 8-bit per channel document, there's more data loss that could even be visible compared to doing the same edit in 16-bits per channel. WHY are all the high bit scanners, cameras, and software just wrong about asking us to work with a file twice as big? WHY is there high bit data for editing then? The answer is to avoid data loss (destructive editing of pixels) with such padded data that the loss is moot, something I said yesterday and a year ago. But the data loss is there. HOW can you call this non destructive editing?

[TheDigitalDog]

>>Note that there is a one-to-one correspondence between the "layer on/off"-eye symbol for a PS adjustment layer and the "on/off" switch for an LR panel.

It appears that way because you fail to understand the difference between a pixel edit and a parametric edit that THEN requires rendering to create pixels. This has nothing to do with destructive or non destructive editing and more to do with the processing workflow, path and engine. Photoshop doesn't render pixels, they are already pixel based. LR and ACR have to take instructions and build pixels.

What do you think happens under the hood in LR when you take your raw data (have you even seen with non demosaiced image data looks like?) a set of instructions and export (setting the rendering) to build, create, compose NEW RGB pixels.

The once amazing "Layers" product that started this thread doesn't do anything differently than Photoshop but it isn't at all working within the LR engine. So you might as well just use Photoshop or any pixel editor that supports layers. The product mentioned first renders the pixels just as if you exported an image in LR to a size and color space from raw data, opened them in PS and added layers. The LR engine is completely out of picture when this plug-in, like all the others that have no access to the LR parametric rendering engine start working. They can't deal with the LR engine, don't understand the parametric instructions etc. The processes are different (and to round trip this discussion, one is destructive).

[areohbee]

I'm sure Adobe could implement non-destructive layers in Lightroom, if they wanted to (I mean without extreme contortion...). C1 has done it, as has Bibble... granted, since Adobe already has layers implemented in Photoshop, they have less incentive than C1 and Bibble did.

Adobe has not designed Lr to readily support true layering, and since they have it already in Photoshop, it may be a while before we see it in Lightroom.

But there is no question in my mind - it is not less suited for non-destructive editors than it is for pixel editors, and I'm sure PhaseOne and (the people who were working at) Bibble Labs would agree, even if some fairly knowledgeable people would not.

[TheDigitalDog]

>>I'm sure Adobe could implement non-destructive layers in Lightroom, if they wanted to.

Agreed! And if memory serves, before LR had selective edits of any kind, C1 and a few raw converters others had this functionality and a bit later we saw this in LR.

[FeedbackCommunityMember]

FYI, I've written my own RAW converter once. Just to play with some ideas.

I initially wrote that I'd stop my attempts at explaining my point of view here, but that was before I saw your latest post in which you asked some question. I hope the answer will help to consolidate the discussion.

[FeedbackCommunityMember]

Andrew, when I wrote my last comment, I didn't see that you had asked some questions in another post.

>> OK, let's back up since we may have a semantics issue here.
We do have a semantics issue because your understanding of "non-destructive" is different from mine.

Your notion of "non-destructive" seems to involve the idea of information preservation. If there is any kind of information loss, you seem to see a case of "destructive editing".

My notion of "non-destructive" means that you can always revisit earlier editing decisions without losing editing effort you put in after making the editing decisions you are revising. This, of course, includes the decision to undo (or deactivate) all editing adjustments you have made to an image, which gives you back the original image.

The original image can always be retrieved when using non-destructive editing (often referred to as "parametric editing) because the original pixels are never changed. All editing is based on the notion of "transformation". A curves adjustment in LR does not change the original pixels, it transforms them into new copies. The same holds true for a curves adjustment layer in PS. It does not destroy the pixels of the layer(s) below it. The latter are untouched and retrievable on demand. It only generates (you called it "create") new pixels that may be transformed by further adjustment layers.

Now the fact that you can export, print, "save as", etc., an image that consists of pixels that have been created by a transformation chain (the chain can be LR's image pipeline or a stack of PS adjustment layers, it doesn't matter) and that by doing this you of course do not have the same information in the target artefact you started with, does not matter at all.

The target artefact (exported image, image print, etc.) has no "non-destructive" properties. However, the transformation process that led to the artefact has. You can always change your mind about transformation steps, including throwing them all away to be left with the original image. You can change your mind about the first transformation step you decided to do (e.g., making a curves adjustment), alter the parameters (here, the curve), and see the effect ripple through all the subsequent transformation steps (e.g., hue adjustments). As a result, the latter transformation steps are still as effective as they were when you created them. In other words, you haven't lost your edit adjustments that followed the initial curves adjustment, even though you have changed your mind about what curve is the best to use. That is the essence of non-destructive editing for me.

>>OK, so just what do you suppose happens if:
I don't flatten but I print the document in Photoshop? The underlying pixels do not get affected? Of course they do.

The original pixels are not affected. PS creates a new rendering from all the layer information and sends that new rendering to the printer. After printing, you can switch off all layers and see your original image. How would that be possible if these pixels were ever "affected" as you write?

>>You have a single pixel with a value of R35/G55/B65 OK. Now I pop an Adjustment layer on top such that the values will be R37/G55/B65. Only ONE value has changed. Your take on the PS processing is, the original values are sent to the printer or the NEW values? If the NEW values are sent, it's a destructive edit due to the rounding errors.

Of course the NEW values are sent, but because the old values are still available, i.e., they have not been destroyed, the editing was non-destructive.

Note that when you use a brush and paint over your R35/G55/B65 pixel so that it is a R37/G55/B65 pixel afterwards then it was a destructive edit. Once you leave PS and the undo history is purged, there is no way you can retrieve the original R35/G55/B65 information. The situation is different if an adjustment layer creates a new (temporary) pixel by transforming the R35/G55/B65 into a R37/G55/B65 pixel, because in this case the R35/G55/B65 pixel is still available to revert to, or to base a different transformation on.

>>WHY are all the high bit scanners, cameras, and software just wrong about asking us to work with a file twice as big? WHY is there high bit data for editing then? The answer is to avoid data loss (destructive editing of pixels) with such padded data that the loss is moot, something I said yesterday and a year ago. But the data loss is there. HOW can you call this non destructive editing?
Even if information is lost in a transformation process, such as going from a 16-bit image to an 8-bit image, the editing paradigm can still be non-destructive if it is always possible to revert to the original 16-bit data. This can be achieved by not overriding the 16-bit data with 8-bit data, but by generating the 8-bit data as a new copy through transforming the 16-bit data into 8-bit data.

All LR editing panels can be thought of setting the parameters of a particular transformation stage in the LR image pipeline.

Every adjustment layer (not regular layer) in PS can be thought of a transformation stage in a transformation chain that is made up by the entirety of all adjustment layers.

This is why PS can (partially) offer a non-destructive editing experience and, conversely, why all what is considered hard core pixel pushing can be incorporated in LR if it is treated as a transformation rather than a (destructive) overriding of data.

[TheDigitalDog]

>>My notion of non-destructive means that you can always revisit earlier editing decisions without losing editing effort you put in after making the editing decisions you are now revising. This of course, includes the decision to undo (or deactivate) all editing adjustment you have made to an image, which gives you back the original, untempered images.

OK, we're getting somewhere in terms of our definitions. So let me ask you. Any application that has a Save As command, or multiple levels of Undo is therefore non destructive? That would mean MacPaint from early 1985 or so was a non destructive editor. The original IS not undergoing destructive edits. To which I'd say so what? .

My take is this is just a recent (within less than a decade or so) marketing hype language if we both agree that any application that can save off an iteration leaving the original alone is supposed to be 'non destructive'. MS Word is a non destructive editor, as are nearly ever application I own. I can always revisit earlier editing decisions without losing editing effort.

My take was, the result of the edit, the data you end up with (because you want that edit) IS destructive due to rounder errors.

The term Non Destructive edit SHOULD tell us about what we just edited, not what didn't get touched, wouldn’t you agree?

>>However, the transformation process that led to the artefact has.

Agreed! The result of the edit may lead to artifacts. But isn't that what's important here? The result of what we do to the image? Didn't the new iteration suffer some data loss? If so, how is this a non destructive edit?

>>The original pixels are not affected. PS creates a new rendering from all the layer information and sends that new rendering to the printer.

Agreed, so the edit isn't non destructive in terms of what ends up being sent to the printer right?

>>Of course the NEW values are sent, but because the old values are still available, i.e., they have not been destroyed, the editing was non-destructive. non-destructive.

I'm concerned with the results of the edit, not what didn't happen to the original. That's why I brought up the differences in high bit (what Photoshop calls 16-bit) and doing the same editing on 8-bit per color channel. I think we both agree there's data loss in both, but one is so highly 'padded' with data, it's moot. But there was data loss. But the 8-bit per color document has undergone data loss to the degree we could see the result of the edit. In this context, why is the original that wasn't touched important, how is this a non destructive workflow?

If I don't edit an image, that too is non destructive in terms of the original data. So what?

Let's now move to LR/ACR. In the marketing hype description, the original raw isn't touched. But how about the resulting rendered data from the parametric instructions? Are we not creating new, virgin, rendered RGB pixles from raw+instructions? Is this not more a real non destructive edit? I'm concerned with the data I receive AFTER an edit, not before.

This is a bit like the question, if a tree falls in the forest and on one hears it, did it make a sound? I say yes. If I edit an image and the original isn't altered because of either a Save As or an adjustment layer, is there data loss? I would say yes (on the edited data). The original isn't touched agreed, but why did we edit the image and what is the result of the image we DID edit?

[areohbee]

If you can't "destroy", or create, or modify pixels - for *output* anyway, then it's not much of an editor.

My .02:

Whether editing is destructive or not depends solely on whether you can go backwards or not.

If you can't go backwards anymore, then your edits have been destructive.
If you can go back to where you were before, then your edits have been non-destructive.

Granted, the manner in which one may be able to go backward, in Photoshop, is radically different than the manner in which one goes backward in Lightroom, and therefore what's easy/hard to do non-destructively can vary a great deal. Still, it seems to me that the definition of the term "non-destructive" is pretty darn straight-forward.

PS - I hope I don't get sucked too far into this discussion, but I guess it's the risk I take by chiming in...

Disclaimer: I have most definitely not read everything written recently, word-for-word, so please forgive (and ignore) if my comments really don't fit into the debate.

Rob

[TheDigitalDog]

>If you can't go backwards anymore, then your edits have been destructive.
If you can go back to where you were before, then your edits have been non-destructive.

And nearly every application on this Mac can do this as I pointed out (Save As, Undo etc). MS word has even more provisions than that, so it's a non destructive text editor. I asked, so what? There's nothing new here.

That fact now brings us to the effect of the edit of our data. We both know the theoretical benefits of high bit capture and editing and why it is provided to us.

IF you send me the best quality capture you ever made and I size it way down, add too much noise and use the Posterize command then save as a JPEG at quality 20, what's the effect on THIS data? That the original is untouched doesn't affect what I just did a one bit.

My take, and I'm trying to be open to the concepts and semantics here, is that what happens to the data we DO EDIT is what should be discussed, not what DID NOT happen to the original that was never edited.

I think TK and I are in agreement about what happens to the data sent to a printer with Adjustment layers (there is some destructive data loss. Actually I'd prefer to just say data loss as destructive sounds serious. It certainly can be).

What we need to agree or disagree upon is the result of taking raw data and a set of instructions to render RGB pixels. In terms of that data, not the raw original we can't view anyway, is this destructive? Forget the original. Let's concentrate on what we end up with after we do an edit. One way to edit is build instructions for rendering our raw data, the other is editing those numbers afterwards in Photoshop with or without Adjustment layers.

Again, calling a process where an original document isn't touched at all a non destructive workflow, while ignoring what happens to the data we edited for a reason seems like a huge exercise in marketing speak. Or am I really missing something?

[areohbee]

Yep - undo and save-as are good to have if you are using a destructive editor, but may not be a sufficiently satisfying substitute for a non-destructive editor.

[TheDigitalDog]

>undo and save-as are good to have if you are using a destructive editor, but are not really a substitute for a non-destructive editor.

I don't follow you. If you undo, no damage. Just like if you Save As... no damage to the original. So are not both processes make the application non destructive?

[hillrg]

Again, calling a process where an original document isn't touched at all a non destructive workflow, while ignoring what happens to the data we edited for a reason seems like a huge exercise in marketing speak. Or am I really missing something?

I don't think anyone would disagree with you Andrew that some edits can be less than optimal. It is just that the original discussion was around the premise that some operations are not possible or reasonable to execute in a parametric editor, not the quality of each edit.

I think we got sidetracked on the term "destructive".

As TK has stated, I see a lot of user requests being arbitrarily dismissed because they need "pixel pushing" and can't be done in a "parametric" editor. We all understand that there are performance and logic issues in the parametric model.

[areohbee]

Undo is limited. For example, in EditPadPro, although undo persists through saving/restarting, the number of steps is limited. In Lightroom, the number of steps is virtually unlimited, but can't be confined to a single photo, and is cleared upon exit.

Save-as leaves a trail of bread-crumb files, and I could say more but shall refrain.

So, in my mind, neither undo nor save-as are equivalent to non-destructive editing, at least not for the long haul...

[areohbee]

"I see a lot of user requests being arbitrarily dismissed because they need "pixel pushing" and can't be done in a "parametric" editor."

- paraphrased by Rob:

"That's already in Photoshop, and may not be easy enough to do given Lightroom's design".

R

[TheDigitalDog]

>Undo is limited.

Yes it is. No question 5 levels of undo is far more useful than 1. But how is 1 undo or a Save As not accomplishing exactly what the so called non destructive workflow describes (we didn't apply a destructive edit to the original data)?

>So, in my mind, neither undo nor save-as are equivalent to non-destructive editing, at least not for the long haul...

That is leading me down the semantic rabbit hole, sorry. I thought the definition of 'non destructive editing' I'm hearing here is that the original is left alone. Which an Undo would accomplish, or a Save As.

I find the entire notion that doing something that doesn't affect the original (despite what may happen to the saved iteration) not a useful concept and not non destructive.