Editing Layer Masks

[cathodeRay]

A lot of useful information here as well as interesting discussion - thanks. My latest PL quirk is the background to the main image view area going magenta, with thumbnails and the main image slow to update, if at all sometimes. May be a memory problem but I should be OK (4GB RAM on XP SP3), or at least I would have thought so.

I'm continuing to experiment and hopefully all will fall into place. At least I'm clearer now about the Alt-Click on the thumbnail (gets access to grey scale editing of the layer mask, same as but quicker than going to the 'special ('secret'?) channel' in Channels. The mask channel when turned on as well or on its own is I suppose the alpha (transparency) channel - but then one has two masks... again, potential for confusion). In fact since it is quicker (and more intuitive - editing the layer mask in place), I wonder whether the special channel functionality actually adds anything?

Will report back when I've made some more progress.

[bkh]

Found that my mask color is black. I see something when I switch to magenta, but still this is not an enlarged version of the thumbnail.

Oops, I had switched mask display off (in the toolbar), and it worked as expected. If I turn mask display on and set the mask colour to magenta, the screen output looks like yours. Stranger still, if I then do "Deselect", the image turns black. Probably a bug, imo, the mask display shouldn't have an influence on the channel display, and one should be able to use the mask normally when editing the special channel.

Cheers

Burkhard.

[Martin Huber]

Oops, I had switched mask display off (in the toolbar), and it worked as expected. If I turn mask display on and set the mask colour to magenta, the screen output looks like yours. (...) Probably a bug, imo, the mask display shouldn't have an influence on the channel display, and one should be able to use the mask normally when editing the special channel.

I don't know, why the mask display influences the special channel. I will change that.

Stranger still, if I then do "Deselect", the image turns black.

That's not really strange. If you turn on the special channel by Alt+click on the thumbnail, the "normal" channels are turned off.
Now if you deselect the active layer, there is no longer a special channel, so all channels are turned off and nothing is drawn. As a result you get a black window.

Martin

[cathodeRay]

Still not got it working, I'm afraid. Test image is a simple red background with a hue adjustment layer adjusting the red to orange. In the first screengrab I have added three adjustment layer masks as per Ken's 'let's keep it simple' (!) instructions: in contrast to Ken's screengrab, only the top mask has any effect. So far as I can see the setting are identical, and the only difference is my expand/collapse icon immediately to the left of the AL thumbnail points SE rather than South. (I've only included the layer stack to keep the image file size down but the main image reflects what the top layer shows).

sibling masks.jpg

The next screengrab is Herbert's schema with the sibs grouped. Even when the blend mode is changed to multiply (and %s changed), nothing gets through. Toggling various visibility/mask/clipping states means sometimes something gets through, but never the full 123:

sibs grouped.jpg

To get the lower grouped masks I have to use (and it sort of makes sense given what the blend modes do) either screen, lighten or difference mode:

sibs grouped on steroids.jpg

All a bit of a palaver. What I was hoping for was for version one (top screengrab above) to work ie the effect of sibling child layer masks would simply be additive. The reason for not having all the mask on one mask layer being that sometimes it is simpler to keep separate parts of a complex mask, err, separate.

[Hoogo]

For #1 I see that there's an additional eye in the check box in the layer list. That's a mode where the mask should be shown instead of the layer it belongs to, I'd call it a kind of "debug mode". Behaves somewhat odd when the parent layer is an adjustment layer, as the adjustment is still applied.

I quickly created some masks for the curves layer, and the masks appear without the eye, see attached file.

[Martin Huber]

in contrast to Ken's screengrab, only the top mask has any effect. So far as I can see the setting are identical, and the only difference is my expand/collapse icon immediately to the left of the AL thumbnail points SE rather than South. (I've only included the layer stack to keep the image file size down but the main image reflects what the top layer shows).

sibling masks.jpg

As Hoogo already said: additionally to the clipping property the layers 1, 2 and 3 are visible. Usually you don't want that.
If you turn off visibility of these layers, their effect will be the intersection of the areas, that they allow to pass through (the non-black areas). Because there is no intersection, the adjustment layer will have no effect.

The next screengrab is Herbert's schema with the sibs grouped. Even when the blend mode is changed to multiply (and %s changed), nothing gets through. Toggling various visibility/mask/clipping states means sometimes something gets through, but never the full 123:

sibs grouped.jpg

If you use Multiply for layer 1 and 2 (layer 3 is the "background" layer of the clipping group and doesn't need a special blend mode), you get exactly the same result as in you first test with visibility turned off: the intersection of the three layers.

To get the lower grouped masks I have to use (and it sort of makes sense given what the blend modes do) either screen, lighten or difference mode:

or "Linear Dodge". And there are probably others.

All a bit of a palaver. What I was hoping for was for version one (top screengrab above) to work ie the effect of sibling child layer masks would simply be additive.

Sibling child layer masks are always subtractive.
If you want to have an additive behavior, you will have to use a clipping group as in your last example. And a clipping group behaves just like a normal layer stack: top layers may hide the bottom ones, so you either have to use blend modes or transparency to let the background shine through.

Martin

[cathodeRay]

BTB - why do I want additive multiple masks? Say I have a luminosity mask/adjustment. It does the global adjustment but I then want to dodge some faces that are a little too in shadow. Rather than (destructively) dodge on the luminosity mask, it would be more flexible to add the face dodging on a separate layer mask. If I blow it, no big deal, just paint back some black/start over - all with the global luminosity mask still intact.

Hoogo - Ken uses a black theme and I mistook the state of the icons - but even when I turn off the eye I still don't get a additive mask:

sibs eyes off.jpg

Even odder (! - maybe you are doing something different, well, you must be), when I open your masks image, it behaves correctly:

additive masks.jpg

Martin -

If you turn off visibility of these layers, their effect will be the intersection of the areas, that they allow to pass through (the non-black areas). Because there is no intersection, the adjustment layer will have no effect.

I think this is the 'and' rather than 'and/or' (either/any mask) effect I think I mentioned earlier, and sort of confirmed by your subsequent remarks:

Sibling child layer masks are always subtractive. If you want to have an additive behavior, you will have to use a clipping group as in your last example. And a clipping group behaves just like a normal layer stack: top layers may hide the bottom ones, so you either have to use blend modes or transparency to let the background shine through.

This is what I need to get my head round (to achieve the sort of usability I described at the top of this post).

Again, thank you everyone for your input - very much appreciated.

cathodeRay

[Hoogo]

Even odder (! - maybe you are doing something different, well, you must be), when I open your masks image, it behaves correctly:

No, it's the same, my mask is also subtractive, the dark side wins.

[cathodeRay]

I was worried about the dark side...

Actually, I realised as I went through my last post that the explanation was down to subtraction, or what I confusingly called 'and' mode ie the white has to be in both masks (in mask one and mask two) to get through.

cathodeRay

[photoken]

BTB - why do I want additive multiple masks? Say I have a luminosity mask/adjustment. It does the global adjustment but I then want to dodge some faces that are a little too in shadow. Rather than (destructively) dodge on the luminosity mask, it would be more flexible to add the face dodging on a separate layer mask. If I blow it, no big deal, just paint back some black/start over - all with the global luminosity mask still intact.

Sure, but are you forgetting that, when using masks, "White reveals, black conceals"?

Here's your desired result following my previous steps:

bb adjustment comparison.jpg

bb adjustment layers.jpg

Added:
I was composing this message before you posted your last message, but what the heck, might as well include the photos....

[Herbert123]

So we have two basic additive mask workflows: with or without transparency. And the background layer does not require to be transparent, btw.

Correct?

Untitled2.png

Untitled.png

[bkh]

So we have two basic additive mask workflows: with or without transparency. And the background layer does not require to be transparent, btw.

Correct?

I'd rather use "Screen" or "Linear Dodge (Add)" than "Difference" because "Difference" produces an exclusive-or (overlapping white turns black again). "Lighter" or "Lighter Colour" will also work. There are slight differences between the approaches when using overlapping soft masks, but as long as all the masked areas are disjoint, the results will all be the same.

Cheers

Burkhard.

[Martin Huber]

So we have two basic additive mask workflows: with or without transparency. And the background layer does not require to be transparent, btw.

Correct?

Yes. In fact, it is more efficient, if the background doesn't have transparency.

Martin

[cathodeRay]

Ken - OK, I have got an exact working replica of your stack, only my background is at the bottom of the stack. The two masks (I used a LAB luminosity mask as the lower mask, painted the upper mask) interact as expected and from what I can see are doing so additively: the darker the combined mask (visible in the in the internal curves AL mask), the less the horror curve is applied. I'm still not clear why me earlier attempts didn't achieve this, but I suppose the answer to that question belongs to 'nice to know' rather than 'need to know'.

Martin, Burkhard & Herbert - I now appear to be able to get additive masks using painting on transparent mask mask layers without grouping and with normal blend mode. The AL in the screengrab below is a full desaturation, lower brightness adjustment layer:

transparency masks.jpg

Where the two masks intersect at the centre of the X the mask is darker, the protective effect (against the desat layer) greater.

If I then turn off the transparency for the mask layers and change the blend mode for the upper mask to screen only the centre of the X shows at a lower intensity (lighter grey) but if the blend mode is multiply then I get (exactly? - need to do some eye-dropper tests) the same effect as with the transparency masks:

greyscale masks.jpg

All of these methods (Ken's, grey/black painting on transparency and greyscale+multiply blend mode) now seem to work, at least for now!

cathodeRay

[Hoogo]

Martin, Burkhard & Herbert - I now appear to be able to get additive masks using painting on transparent mask mask layers without grouping and with normal blend mode.

No, still subtractive Black=0, White=1, and every mask removes a little more, it gets darker and darker, less work is done. Maybe you are used to CMYK, there the colors are meant the other way around?