Passthrough mode blend breaks combined with child layer masks or erase mode

[CM0]

In my continuing efforts to break down and document the sometimes odd Affinity behavior, I'm now focusing on the peculiar blending behavior of passthrough mode.

Previous explanations have been something like "it is due to blending order"

However, if this is the case, we should be able to manually replicate the exact same behavior. I have been completely unable to do so. I can not replicate, nor explain it. Furthermore, the behavior would not seem to be expected or desirable.

This is a sample using passthrough

And this is the same using normal

There are specifically 2 problems (bugs) with the previous image:

1. Colors are deformed. There seems to be desaturation or some additional oddities. Pure red blends normally. However, anything with mixed colors or partial grays results in undesirable color transformation.
2. There is banding in the image. It seems the underlying smooth gradient is somehow clipped and is no longer a smooth ramp. This is most evident in the middle gray layer. If you sample the colors, you will see the values of gray do not change at all until you cross the midpoint of the image.

These conditions only occur when we have a mask or erase mode layer like the following:

The mask also must have partial transparency. It is the area of partial transparency that produces the unexpected results. But this must be done with a mask, as if we place an image with partial transparency it will blend fine.

I had assumed this is likely due to the mask being applied to the underlying layer ( in the above would be fill layer ) as part of blending/rendering and then blending a second time as part of the group with the fill again. However, you can't reproduce this effect by manually doing these steps. No such combination of doing that with any blend mode available in affinity will result in what we observe with passthrough. No matter how you blend with either the black fill or an empty layer or multiple merges results in what we observe. Additionally, if you add the mask as a child layer of a child, the problem still occurs and in that configuration the mask should not be affecting the group's underlying layer.

So then what is happening? I've attached the sample document for anyone that would like to attempt to further explain.

What to look for and when to use work arounds:

Anytime you are using passthrough combined with some type of applied transparency (mask, erase mode) you likely will not get the expected result. Consider ...

1. Don't use passthrough. Break up your composition in some other way. This might be difficult depending on your composition.
2. Create copies of your mask and apply to every layer individually that is a child of the group layer. The mask must be attached to the masking position. If added as a child layer of a child layer the problem will resurface. This is another odd behavior.

 

passthrough mode blending.afpub

[lepr]

On 9/18/2023 at 3:59 PM, CM0 said:

Previous explanations have been something like "it is due to blending order".

However, if this is the case, we should be able to manually replicate the exact same behavior. I have been completely unable to do so. I can not replicate, nor explain it. 

I gave you the blending order explanation for a particular scenario we discussed. There was no partial transparency involved in that scenario, and my explanation was correct for that scenario. Your new example which does involve partial transparency is a different matter, and I even remarked that masked groups was something to be discussed at some other time.

[CM0]

6 minutes ago, lepr said:

I gave you the blending order explanation for a particular scenario we discussed. 

This wasn't about you. Do you have something you would like to add here that is relevant?

[lepr]

There is something relevant I could add regarding the mathematics of what is happening, but I haven't yet decided whether it would be worth the effort to write it up since the developers' roadmap isn't going to be influenced by it.

[CM0]

1 minute ago, lepr said:

 since the developers' roadmap isn't going to be influenced by it.

It would seem this is in reality the most critical issue above all others. :-(

Sometimes I don't even know why I'm here. I think exactly zero bugs have been fixed of all those I've submitted over the years. I really wish we could have a release that is just catchup on all the issues.

[Callum]

Hi All,

I have logged this with our developers to be improved in a future update

Thanks
C

[Affinity Info Bot]

The issue "Passthrough Blend Mode with grouped mask is giving unexpected results" (REF: AF-399) has been fixed by the developers in the latest beta build (2.6.0.2805). The fix is planned for inclusion in the next customer release.
Customer beta builds are announced here and you can participate by following these instructions.
If you still experience this problem once you are using that build version (or later) please reply to this thread including @Affinity Info Bot to notify us.

[CM0]

Please note this is not fixed for erase mode that I documented above.

If you use erase mode as a mask, this problem still occurs in passthrough mode.