Are you sure you are choosing Linear Light and not Vivid Light in each case? and are you sure you are viewing the image at 100%? Are you also sure that the layer and FILL opacities of both layer1 and group1 are 100%?
PS: Vivid light is not the only mode that produces different results with the 2 different methods above. This also holds true for (at least): Color Dodge, Hard Mix and Difference.
Ok, You are right.
Try this – take example 2 where you have the vivid light layer on its own – choose the foreground(lights) layer, go to Edit/Fill .. chose "behind" from the drop down, and fill this using 50% grey – see its transformed into the same effect you are getting with the group?
Its explainable in this way .. layer groups when you apply a blend mode to them take their VISIBLE contents and apply the mode to the whole area, so its similar to applying a neutral 50% grey behind transparent pixels.
THEY DO behave differently but that gives more flexibilty
Thanks so much for trying that out with me Mark. Apparently you learned something, too, given that you thought my OP was somehow wrong ? (or did I explain things just too poorly? probably …)
Your explanation is very logical indeed. But is this behaviour documented anywhere ?? That’s the first time that I’m hearing of that. If it’s a undocumented feature, how are users supposed to find out? I mean, who the hell would put a layer into a group (and what is more, being the only layer there!), and change the group’s blend mode instead of the layer’s one, if there are no other layers to interact with ?
Also notice that the 50%-grey-fill does not *exactly* reproduce example 2 above. Make snapshots of both and go back and forth between them at 400% and see the differences. They grey-fill-version looks noisier at that magnification level. Any ideas why? Following your explanation – if groups do exactly the same as we did manually with the 50% grey fill, the results should be the same. The differences are tolerable, but the inexactness puts the whole concept into question.
Here is a section of the image. I difference-blended the two methods to show you the changing pixels:
What’s interesting, on the other hand I compared the color dodge blending with both methods and all resulting pixels are 100% identical. Strange …
Also, to better understand how groups are really behaving: You said, they apply a neutral 50% grey behind transparent pixels. Now that might well be the case with a group in vivid-light-mode. But what about modes where there is no neutral colour, like "hard mix" for instance ? How do groups behave then ?
PS: The problem with that group-thing is also that I can’t use the layer as a clipping mask anymore, since it’s in a group now and groups cannot act as clipping mask, right?
You can still clip with other layers WITHIN the group, or you change the group into a smart object and clip to that.
My theory as to the slight difference in the effect (and it IS very slight, difference and threshold bring up ) may be something to do with grey being neutral in Vivid Light. In some of the other modes, black or white are the neutrals. I think some kind of inconsistency created in one case using the document profile and in another case not using it? Just a theory
The I think you are pulling apart and spotting some interesting oversights. Hard mix behaves completely differently to the way I would expect it to within a Hard mix group. Maybe these are genuine oversights and bugs. There has historically been an issue with certain blend modes shifting from release to release meaning that older files opened wrongly.
Another place you see these inconsistencies exist is with Layer styles – try some of these modes on a layer style and then see what happens when you create layers
I think you have brought up some interesting things here, and it would be really interesting to have some other input.
Hm. Possible theory. Is there a place to file potential bugs and get some dev feedback?
"Within" the group : yes. But I need to act the layer as a clipping mask for a layer outside the group.
And "change the group into a smart object": It doesn’t work. I loose the blending effect and get back the white halos (from the second screenshot above). So all in all, this "group-thing" is rather annoying and restricts you rather than give you more flexibility. Or do you see any solution ?
Sorry for asking again, but that’s paramount for me: You said, before blending groups fill transparent pixels with the blend mode’s neutral colour. I never heard of that. Is that documented somewhere? And what about groups with a blend mode for which a neutral colour does not exist ?
For anybody who wants to reproduce the issue, you don’t need to DL the sample files above. Just use any appropriate background and a white text layer instead.
I certainly didn’t say "group-fills-transparent-pixels-with-neutral-color", I was just trying to illustrate how photoshop calculates blend modes within a group – The "fill behind" example I gave you was to try and illustrate this only. Also "You said, before blending groups fill transparent pixels with the blend mode’s neutral colour" – no I didn’t say that, the fill behind thing is just to show how its possible to reproduce the effect in a different way.
I don’t understand your last paragraph.
Ok, sorry for misinterpreting you then, I thought that was what you meant when saying: "layer groups when you apply a blend mode to them take their VISIBLE contents and apply the mode to the whole area, so its similar to applying a neutral 50% grey behind transparent pixels."
So if that’s not what you meant, than I’m basically back at square one and don’t know at all, HOW you explain what accounts for the different blending results of a layer vs. of a group.
All livedocs has to say on this is: " By default, the blending mode of a layer group is Pass Through, which means that the group has no blending properties of its own. When you choose a different blending mode for a group, you effectively change the order in which the image components are put together. All of the layers in the group are put together first. The composite group is then treated as a single image and blended with the rest of the image using the selected blending mode. Thus, if you choose a blending mode other than Pass Through for the group, none of the adjustment layers or layer blending modes inside the group will apply to layers outside the group."
I couldn’t find any other "official" definitions of the way groups affect blending.
"…is then treated as a single image…" -> so basically there shouldn’t be any difference at all between my two screenshots above. They should blend the same. A different result is against livedoc’s definition.
What I meant with my last paragraph is that the "blend layer" used in my example above contains no (or almost no) 100% opaque pixels. So it’s a particular case and thus maybe not the best starting point for evaluating basic rules. That’s why I suggested to just use a white text layer instead (which has 100% opaque pixels).
Try the same thing as above: Cycle through all the blending modes for the layer. Then put the layer inside a group and cycle through the group’s blending modes. You’ll see, the results are not the same, and again, that’s against the abovementioned definition.
There must be other rules in play as well here. I did an extensive Internet search, but couldn’t find any other rules, that would explain the results.
No offense, but nobody here knows how a layer group’s blend modes work ?
We do, but we’re not telling.B)
This would be the occasion for all the folks who criticize me so heavily (and probably think that’s the purpose and main reason for this forum’s existence) for allegedly not understanding transparency and blend modes, to show up and do something positive/constructive instead of mere fingerpointing and engaging in "I’m better than you, you don’t understand anything"-philosophies, because on this issue I readily admit (and have consistently done so in this thread) that I don’t understand what’s going on. I’d be very, very thankful for anybody explaining the abovementioned behaviour. I searched a lot on the net and couldn’t find any documentation on it, so I’m sure there are lots of users out there who would find an explanation equally useful for them too.
