NotMyFault
-
Posts
6,692 -
Joined
-
Last visited
Posts posted by NotMyFault
-
-
- Duplicate layer
- add mask from selection of shirt
- deselect all
- add curves adjustment and move right node to brighten shirt to intended lightness of new color
- use color panel to assign color (if layer is image layer), or use hsl / recolor adjustments
-
-
ah partially my fault. for unknown reasons, the PT filter has inherent mask with partial alpha. Probably happened by mistake when I adjusted a color layer.
Never the less, why system HDR settings make a difference is mysterious to me
-
The Problem seems to be bound to Display being in HDR/EDR mode in system settings.
After deactivating HDR mode, issue vanished.
Never the less, this should be fixed as Affinity Apps are advertised as HDR/ERD capable.
-
The same file works flawlessly on iPad. So I assume a bug with Mac version.
-
I tested the same process using a channel mixer adjustment:
R=R+B
G=G+B
B=B
and duplicated the channel mixer many times. this works with any amount of layers.
Why does the PT filter gives different results than the channel mixer adjustment when using essentially the same formula?
-
found a strange issue.
Using a PT filter to add a small amount to selected color channels, the PT filter stops working after 32 copies.
Basic principle:
base fill layer gives R and G as zero, and B as 1 (of 255)
PT filter adds value of B to R and G channel. B values stays constant.
Use info panel to inspect color values (and your eyes to see canvas).
The filter adds up nicely until color value 32. Then it stops to add.
You may modify the base layer color values, e.g change R value from zero to 27 and result still stays at 32. So it is not depending on number of PT filters, it is depending on color values. Only when setting R to 32 you start to get higher values (after pT filter)
I use other file with hundreds of PT filters working fine.
Mac mini M1 (2020)
Affinity 2.3.1 retail and beta 2.4.0
-
57 minutes ago, Scripto said:
Not my fault (great name ... I'm assuming you had many occasions to use it?)
Yes 👍🏼
except when I'm in err - happens only 41 times. day.
-
56 minutes ago, Scripto said:
Thank you for your solution but I have three questions: 1. won't that leave a white halo around the object?
no. It will perfectly blend. Did you try yourself? You don't need to trust my words
56 minutes ago, Scripto said:2. What about a freeform object (not symmetrical geometry?)
Then you need a different approach.
- Duplicate the freeform shape. (duplicate linked to ensure it will update later you adjust the shape)
- Add gaussian blur, nested to masking position of duplicate
- nest the duplicate to masking position of original layer.
- adjust blur radius as needed.
56 minutes ago, Scripto said:, but if this program has a function that doesn't work, shouldn't it be fixed?
Yes I totally agree. But I'm only a paying customer, not Affinity / Serif staff.
-
feather selection is not the best approach in this specific case.
You always need to convert this selection into a mask, or use copy/paste, not allowing the easily adjust the radius later.
If you want a circular shape with feathered edge:
- add a circle
- add a circular gradient
- set both points to white
- add 3rd node near edge where you want the feather to start
- select node at edge
- reduce color opacity of that node to zero
- Nest this shape into masking position of your pixel layer
- adjust the gradient node position to taste / as needed
-
-
Another solution which works better in certain situations:
- Duplicate the layer you want to feather
- nest one to the other in masking position
- add live Gaussian blur in masking position of inner
- Adjust strength to taste
- in my experiments the alpha values did not reach 0 to 100 but only a subset of this range. To correct:
- add a channels mixer above child layers
- choose alpha channel
- set offset to -100%
- set alpha input to 200%
now the feather works as wanted.
-
Just found a better solution:
- Duplicate the layer you want to feather
- nest one to the other in masking position
- add live Gaussian blur in masking position of inner
- Adjust strength to taste
- in my experiments the alpha values did not reach 0 to 100 but only a subset of this range. To correct:
- add a channels mixer above child layers
- choose alpha channel
- set offset to -100%
- set alpha input to 200%
now the feather works as wanted.
The file shows a 1px checkered board above a blue fill layer. You can feather the shape from hard edge to soft
- languidcorpse, loukash and GripsholmLion
-
1
-
2
-
You can add a Gaussian blur filter on any shape. This works well for shapes with solid fill colors.
In case of shapes filled with bitmap fills or fine details, you need to restrict the blur filter to the alpha channel, based on the tutorial below:
-
To explain it differently: even if you use a very light color with 98% and assign it to a linked PNG, the luminosity of the result will be that of the PNG, 74% in this case.
to get perfect results, you need to use e.g. an hsl adjustment to correct the luminosity first, before changing the hue/saturation.
-
Your source image luminosity is 74%
your chosen color F4FFF6 luminosity is 98%
so both will match only at 98% and not 74%
-
I tested both V1 and v2 and got identical results.
When you assign a color to an embedded or linked file, the color will only be identical in RGB values if the luminosity of the embedded image is identical.
For a test,
- create a rectangle of 256px
- and use a black tom white gradient.
- Export as PNG
- create a new document
- embed the PNG
- add a rectangle 256px x 80 px
- arrange both to slightly overlap in position vertically
- group both layers
- now use the color panel to assign various colors. This will impact both layers simultaniously.
you will see that colors match in exactly in one of the 256 columns (if at all).
So if you have a png with uniform color and luminosity, you cannot get one specific color, the range of colors is limited by the luminosity of your source image.
you can use other methods to get the desired results.
-
2 hours ago, pfi said:
When weird stuff happens again, I will try to remember and do your suggestions. - thanks
Thanks for your reply.
Most important: in case you want to file a bug report, always attach an afphoto file where having this issue. Filing a bug report and not saving the actual file can cause waist of time for all who try to work on the issue, as nobody will be able to reproduce.
2 hours ago, pfi said:run a macro, which makes a copy of the background image and ads my most often used adjustments and filters, most unmodified, some with minor pre-adjustments.
That is important.
-
Can you provide example files?
Please check that you are using same size (pixel x/y. DPI) and same zoom factor.
-
-
It’s a feature discrepancy. Not available on iPad.
-
4 hours ago, walt.farrell said:
I've been curious about that recommendation. Often, the lower layer will be the background layer, and by definition it has the proper DPI and is aligned to the grid. So why would rasterizing it help?
you are safe when opening a file via „open“.
In case the layer is created by „place“ or „pasted“ into an exiting document it is often misaligned, or will be enlarged or reduced.- Old Bruce, hawk and walt.farrell
-
1
-
2
-
Please keep in mind to check layer size, position, rotation etc. you should add one extra step and rasterize the lower layer before merging them in case of:
- fractional position
- Layer DPI not matching documnet DPI
- layer rotated or sheared.
otherwise result of merge will be blurry.
so in general to be safe:
- rasterize lower layer
- group both layers
- rasterize group
-
Changing color of a shirt on a model/person
in Pre-V2 Archive of Affinity on iPad Questions
Posted