Live Filter (Gaussian Blur) - Doesn't Behave like Standard Filter (or when used in a mask)

[ShawnG]

Affinity Photo v2.5.6 (MacOS 15.1.1)

The Live Gaussian Blur Filter doesn't give the expected results for this case:

1) Start with a new image, fill the background to be solid black.

2) Use the paint brush tool to create an area in the center that is solid white. 

3) If you apply a Filters > Blur -> Gaussian Blur of about 100 pixels.  This works correctly, the white area is blurred and the outer edges of the image are still black.

4) Undo the filter.

5) Add a Live Filter (Gaussian Blur) and set the radius to 100.   The results are different!  The inner white area is blurred, but now the edges of the photo now has a white fringe.     This seems to be a bug as the results should be the same.

 

I can select the "preserve alpha" option and this works for just this case.  But this does NOT work for the case where this type of image is used for a mask -- as the "preserve alpha" option when applied to a mask makes the blur not work at all for the mask (since the mask is essentially adding a alpha-like area).   It is possible the bug is here instead: when applying the "preserve alpha" in a mask for a live filter.

---

This use-case comes from trying to apply masks to layer effects, where most of the mask is black and there are just small areas desired to apply the effect.   I don't want to use the destructive Gaussian blur because I do want to go back and edit the mask area often -- which is the point of live filters.  Every time I do this the border of my image gets a dark vignette effect which is unwanted.

-------

 

 

 

 

 

 

Edited December 9, 2024 by ShawnG
grammar corrections

[ShawnG]

Some screenshots to help with the more advanced case:  Here is the live Gaussian blur applied (without Preserve Alpha) on a Mask of an Curves Adjustment Layer:  Notice the darkened edges around the outer border of the image (compared to the next screenshot) due to the Live filter making the edges white.

[ShawnG]

Here is the screenshot for the same scenario but with "Preserve Alpha" for the Live Filter turned on.  The Blur on the mask is completely gone -- but at least the edges are not feathered.  

[Chris B]

Hey @ShawnG

Sorry for the delay. This has been a bit of a headscratcher as I see your issue but it was a bit tricky finding a potential workaround for you.

What you could try doing is the following:

1. Open your image
2. Add a Fill layer > make it black and rasterise
3. Draw your white area with the Paint Brush Tool
4. Live Gaussian Blur and check Preserve Alpha
5. Marquee select around the white area
6. Invert it (cmd+I)
7. Edit > Fill > Apply

I've tried making a macro for you as well which you can import. There's a Grow / shrink section which you can edit to affect the size depending on your image.

I understand if this is a pain or unsuitable but as of right now, it's the only workaround I can suggest without a change to the way filters and alpha work.

Gaussian Blur macro.afmacro

[lepr]

@ShawnG You will benefit from Affinity's Empty Mask type of raster mask in your scenario, instead of filling the usual Mask type of raster mask with 'black'.

An Empty Mask is an infinite extent of virtual 'black' in which actual pixels can be painted/filled, whereas a regular Mask is an infinite extent of virtual 'white' in which actual pixels can be painted/filled.

Filling a regular Mask with 'black' makes it 'black' only within the boundary of the canvas, but it is considered to be 'white' outside of the canvas, and so partial transparency results at the edges of the canvas if that Mask gets blurred. An Empty Mask is 'black' both inside and outside the canvas boundary, therefore partial transparency does not develop at the edges of the canvas when the Empty Mask gets blurred.

 

[NotMyFault]

On 1/15/2025 at 2:16 PM, lepr said:

An Empty Mask is an infinite extent of virtual 'black' in which actual pixels can be painted/filled, whereas a regular Mask is an infinite extent of virtual 'white' in which actual pixels can be painted/filled.

I can’t follow this, both masks (regular and black) are only defined within the canvas area, there is no „infinite“ extension beyond/outside of the canvas.

When you use the move tool and reduce mask layer size or rotate, the edges always filled with black.

To my knowledge and tests. Affinity only stores the „used“ non-black areas in any case. Both masks only differ by being filled or not filled in the visible canvas, and do not differ wrt the area outside the canvas. Once you paint in the first pixel into the black mask, there is no difference to removing (deleting) all other areas from a white mask.

Edit: further testing showed that @lepr is correct wrt empty mask in theory extending „black“, e.g. when reducing the mask layer . A bug of Affinity Apps on iPad distracted me and lead to a wrong conclusion from my side. I will raise a separate bug report. Nevertheless i don’t like his snooty style of communication with other forum members. Despite my chosen forum alias i have no problem to admit if i was wrong and correct false statements.

[lepr]

7 minutes ago, NotMyFault said:

I can’t follow this [...]

That's OK. The apps can be used without an understanding of the technicalities.

 

 

[NotMyFault]

Just now, lepr said:

That's OK. The apps can be used without an understanding of the technicalities.

 

 

Come on. That sentence is snooty. I was extremely polite expressing that your description does not match what is observable.  If your theory can’t be proved by measuring in the reality it’s wrong.

 

[lepr]

1 minute ago, NotMyFault said:

Come on. That sentence is snooty. I was extremely polite expressing that your description does not match what is observable.  If your theory can’t be proved by measuring in the reality it’s wrong.

My next snooty response: consider the reliability of your testing and your interpretation of its results.

[NotMyFault]

8 minutes ago, lepr said:

My next snooty response: consider the reliability of your testing and your interpretation of its results.

Repeating false claims without any proof and insulting other forum users - what’s next?

[lepr]

Just now, NotMyFault said:

Repeating false claims without any proof and insulting other forum users - what’s next?

3 words: glass, houses, stones

I'm under no obligation to "prove" anything. We disagree about technicalities of the masks, and that's OK.

 

[NotMyFault]

12 minutes ago, lepr said:

3 words: glass, houses, stones

I'm under no obligation to "prove" anything. We disagree about technicalities of the masks, and that's OK.

 

Disagreement over technical topics is ok, switching without reason from professional discussion to personal level and allegations of incompetence is not ok.

i did not pick up any stones or throw anything on personal level. I politely said where I disagreed on technical level, in a very non-aggressive form „I cannot follow“ which intentionally leaves open if the err is on my or your side.

[ShawnG]

@Chris B  thank you for taking the time to look into this and for the potential work around (until maybe it can be address in a future version) and for taking the extra time to even coming up with a possible macro!   I will try to work through those steps when I process my next image (it is unfortunately quite cloudy this time of year).    @lepr Thanks for that idea as well; as it sounds possibly even easier to use a Empty Mask (though that is not what I would have thought they were used for).