rvst

I've tried some of these automated tools and while they work for a good number of images, they don't work all that well on edge cases, which is often when we need them the most. 
There are quite a few different techniques one can use to manually mask out backgrounds depending on what it is you're trying to mask out. 
For difficult daylight sky removals, where there are trees or lots of details on the skyline, I generally use a luminosity mask. This requires multiple steps, so it's not a click and drag, but it gives superior results in my experience. 
In V2 I still do this manually as the new luminosity mask is a live mask and can't be painted on (there are workarounds for using the new luminosity mask, but it ends up no faster than the method below).
Here's my workflow for this particular type of mask
Select --> Tonal Range --> Select shadows Click on Channels panel. Right click on Pixel Selection and choose Create spare channel Now do the same and select midtones instead of shadows In the Channels panel, right click Spare channel and choose Add to pixel selection Now use the use selection brush in subtract mode to deselect areas inside your foreground (and vice-versa if anything in the background is incorrectly selected) Once you have your selection, choose Select --> Grow selection by 1 pixel and then feather selection 4 pixels or less. This will leave a slight halo around your edges if your sky you're masking out is much lighter than the replacement sky, but if you don't do this, then you're likely to get edges that are too sharp and lead to a blocky effect. I never grow by more than one pixel and sometimes feather by less than 4 pixels, never going above 4 (I feathered by 2 on the image below) With the selection still active, click on the mask icon to create a new mask layer from the selection Select the mask layer then alt-click on it to see the black and white mask In sections where the mask should be opaque, paint over it in white if you missed. Ditto for black. Be careful not to go near the edge. We'll deal with that soon. Now click off the layer then reselect the mask layer so you can see the image and not the B&W mask layer Reduce brush opacity to 50% and set brush blend mode to overlay (you can start with even lower opacity and just go over it more times) Zoom in on the border between sky and background and find any areas where there is haloing Now apply the brush (in black) over the edges to reduce the haloing. Make sure you're painting on the mask layer. You may need to increase or reduce brush opacity on the fly. The overlay blend mode is key to getting this step right. You can also apply a curves adjustment to the replacement sky (a U shape, either inverted or not depending on whether you want to reduce or increase the luminosity of the replacement sky - this also helps to eliminate the haloing by reducing the exposure difference between the edges of the foreground and the replacement sky) As a final step, and not always - it depends on the image, I might adjust the blend options for the layer, something like the following. It also helps in reducing the haloing and giving a more natural looking transition. This won't work if you have very strong highlights in the foreground part of the image (as it will then reveal a portion of the replacement image on those strong highlights. The exact shape of the curve would depend on the image.  Needless to say, you should get your image looking right before swapping out the sky, as doing this post swap would likely just accentuate the transition border. 
 
Below is an example image where I used this technique to mask out the sky. First image is a crop from a 400% zoom to show the result, the second is the source image and the last is the final, processed version with the replacement sky added. 

I've tried some of these automated tools and while they work for a good number of images, they don't work all that well on edge cases, which is often when we need them the most. 
There are quite a few different techniques one can use to manually mask out backgrounds depending on what it is you're trying to mask out. 
For difficult daylight sky removals, where there are trees or lots of details on the skyline, I generally use a luminosity mask. This requires multiple steps, so it's not a click and drag, but it gives superior results in my experience. 
In V2 I still do this manually as the new luminosity mask is a live mask and can't be painted on (there are workarounds for using the new luminosity mask, but it ends up no faster than the method below).
Here's my workflow for this particular type of mask
Select --> Tonal Range --> Select shadows Click on Channels panel. Right click on Pixel Selection and choose Create spare channel Now do the same and select midtones instead of shadows In the Channels panel, right click Spare channel and choose Add to pixel selection Now use the use selection brush in subtract mode to deselect areas inside your foreground (and vice-versa if anything in the background is incorrectly selected) Once you have your selection, choose Select --> Grow selection by 1 pixel and then feather selection 4 pixels or less. This will leave a slight halo around your edges if your sky you're masking out is much lighter than the replacement sky, but if you don't do this, then you're likely to get edges that are too sharp and lead to a blocky effect. I never grow by more than one pixel and sometimes feather by less than 4 pixels, never going above 4 (I feathered by 2 on the image below) With the selection still active, click on the mask icon to create a new mask layer from the selection Select the mask layer then alt-click on it to see the black and white mask In sections where the mask should be opaque, paint over it in white if you missed. Ditto for black. Be careful not to go near the edge. We'll deal with that soon. Now click off the layer then reselect the mask layer so you can see the image and not the B&W mask layer Reduce brush opacity to 50% and set brush blend mode to overlay (you can start with even lower opacity and just go over it more times) Zoom in on the border between sky and background and find any areas where there is haloing Now apply the brush (in black) over the edges to reduce the haloing. Make sure you're painting on the mask layer. You may need to increase or reduce brush opacity on the fly. The overlay blend mode is key to getting this step right. You can also apply a curves adjustment to the replacement sky (a U shape, either inverted or not depending on whether you want to reduce or increase the luminosity of the replacement sky - this also helps to eliminate the haloing by reducing the exposure difference between the edges of the foreground and the replacement sky) As a final step, and not always - it depends on the image, I might adjust the blend options for the layer, something like the following. It also helps in reducing the haloing and giving a more natural looking transition. This won't work if you have very strong highlights in the foreground part of the image (as it will then reveal a portion of the replacement image on those strong highlights. The exact shape of the curve would depend on the image.  Needless to say, you should get your image looking right before swapping out the sky, as doing this post swap would likely just accentuate the transition border. 
 
Below is an example image where I used this technique to mask out the sky. First image is a crop from a 400% zoom to show the result, the second is the source image and the last is the final, processed version with the replacement sky added. 

I've tried some of these automated tools and while they work for a good number of images, they don't work all that well on edge cases, which is often when we need them the most. 
There are quite a few different techniques one can use to manually mask out backgrounds depending on what it is you're trying to mask out. 
For difficult daylight sky removals, where there are trees or lots of details on the skyline, I generally use a luminosity mask. This requires multiple steps, so it's not a click and drag, but it gives superior results in my experience. 
In V2 I still do this manually as the new luminosity mask is a live mask and can't be painted on (there are workarounds for using the new luminosity mask, but it ends up no faster than the method below).
Here's my workflow for this particular type of mask
Select --> Tonal Range --> Select shadows Click on Channels panel. Right click on Pixel Selection and choose Create spare channel Now do the same and select midtones instead of shadows In the Channels panel, right click Spare channel and choose Add to pixel selection Now use the use selection brush in subtract mode to deselect areas inside your foreground (and vice-versa if anything in the background is incorrectly selected) Once you have your selection, choose Select --> Grow selection by 1 pixel and then feather selection 4 pixels or less. This will leave a slight halo around your edges if your sky you're masking out is much lighter than the replacement sky, but if you don't do this, then you're likely to get edges that are too sharp and lead to a blocky effect. I never grow by more than one pixel and sometimes feather by less than 4 pixels, never going above 4 (I feathered by 2 on the image below) With the selection still active, click on the mask icon to create a new mask layer from the selection Select the mask layer then alt-click on it to see the black and white mask In sections where the mask should be opaque, paint over it in white if you missed. Ditto for black. Be careful not to go near the edge. We'll deal with that soon. Now click off the layer then reselect the mask layer so you can see the image and not the B&W mask layer Reduce brush opacity to 50% and set brush blend mode to overlay (you can start with even lower opacity and just go over it more times) Zoom in on the border between sky and background and find any areas where there is haloing Now apply the brush (in black) over the edges to reduce the haloing. Make sure you're painting on the mask layer. You may need to increase or reduce brush opacity on the fly. The overlay blend mode is key to getting this step right. You can also apply a curves adjustment to the replacement sky (a U shape, either inverted or not depending on whether you want to reduce or increase the luminosity of the replacement sky - this also helps to eliminate the haloing by reducing the exposure difference between the edges of the foreground and the replacement sky) As a final step, and not always - it depends on the image, I might adjust the blend options for the layer, something like the following. It also helps in reducing the haloing and giving a more natural looking transition. This won't work if you have very strong highlights in the foreground part of the image (as it will then reveal a portion of the replacement image on those strong highlights. The exact shape of the curve would depend on the image.  Needless to say, you should get your image looking right before swapping out the sky, as doing this post swap would likely just accentuate the transition border. 
 
Below is an example image where I used this technique to mask out the sky. First image is a crop from a 400% zoom to show the result, the second is the source image and the last is the final, processed version with the replacement sky added. 

Ahead of the curve? This is literally the only significant piece of software I use that chose to deliberately implement Microsoft's horrible tablet centric interface-in-an-interface and totally break the desktop experience.
It was an appallingly bad decision for numerous reasons, some of which have been highlighted in this thread.
I recommend the following fix that you implement ASAP in a maintenance release: when installing, create a stub executable in C:\Program Files\Affinity\Photo that contains a copy of the icon and invokes the executable stored in C:\Program Files\WindowsApps\<mangled installation folder name>\App\Photo.exe. Rinse-repeat for Designer and Publisher. In this way, the applications will be accessible using the standard desktop interface metaphor of Windows as well, instead of only via their ridiculous Windows App model that nobody I know even uses. 
 

It's unheard of in the software industry for software to have forward compatibility, ie., to be able to open a newer version file in an older version of the software (which is  what we're talking about, not backward compatibility as many people are calling it. Backward compatibility would mean that Affinity V2 can open V1 files, which it can, so it's already backward compatible)
The older software would need all the code for the new file format backported into a specific maintenance release to enable this functionality, and apart from the technical challenge of doing so, there is a disincentive for the developer to do this. 
What Affinity SHOULD do, however, is issue a warning when saving a V1 file using V2 software - "WARNING: once saved with the V2 file format, you won't be able to open it in V1". That way, at least the user will know before executing an operation that cannot be reversed. 
 

It's unheard of in the software industry for software to have forward compatibility, ie., to be able to open a newer version file in an older version of the software (which is  what we're talking about, not backward compatibility as many people are calling it. Backward compatibility would mean that Affinity V2 can open V1 files, which it can, so it's already backward compatible)
The older software would need all the code for the new file format backported into a specific maintenance release to enable this functionality, and apart from the technical challenge of doing so, there is a disincentive for the developer to do this. 
What Affinity SHOULD do, however, is issue a warning when saving a V1 file using V2 software - "WARNING: once saved with the V2 file format, you won't be able to open it in V1". That way, at least the user will know before executing an operation that cannot be reversed. 
 

I've tried some of these automated tools and while they work for a good number of images, they don't work all that well on edge cases, which is often when we need them the most. 
There are quite a few different techniques one can use to manually mask out backgrounds depending on what it is you're trying to mask out. 
For difficult daylight sky removals, where there are trees or lots of details on the skyline, I generally use a luminosity mask. This requires multiple steps, so it's not a click and drag, but it gives superior results in my experience. 
In V2 I still do this manually as the new luminosity mask is a live mask and can't be painted on (there are workarounds for using the new luminosity mask, but it ends up no faster than the method below).
Here's my workflow for this particular type of mask
Select --> Tonal Range --> Select shadows Click on Channels panel. Right click on Pixel Selection and choose Create spare channel Now do the same and select midtones instead of shadows In the Channels panel, right click Spare channel and choose Add to pixel selection Now use the use selection brush in subtract mode to deselect areas inside your foreground (and vice-versa if anything in the background is incorrectly selected) Once you have your selection, choose Select --> Grow selection by 1 pixel and then feather selection 4 pixels or less. This will leave a slight halo around your edges if your sky you're masking out is much lighter than the replacement sky, but if you don't do this, then you're likely to get edges that are too sharp and lead to a blocky effect. I never grow by more than one pixel and sometimes feather by less than 4 pixels, never going above 4 (I feathered by 2 on the image below) With the selection still active, click on the mask icon to create a new mask layer from the selection Select the mask layer then alt-click on it to see the black and white mask In sections where the mask should be opaque, paint over it in white if you missed. Ditto for black. Be careful not to go near the edge. We'll deal with that soon. Now click off the layer then reselect the mask layer so you can see the image and not the B&W mask layer Reduce brush opacity to 50% and set brush blend mode to overlay (you can start with even lower opacity and just go over it more times) Zoom in on the border between sky and background and find any areas where there is haloing Now apply the brush (in black) over the edges to reduce the haloing. Make sure you're painting on the mask layer. You may need to increase or reduce brush opacity on the fly. The overlay blend mode is key to getting this step right. You can also apply a curves adjustment to the replacement sky (a U shape, either inverted or not depending on whether you want to reduce or increase the luminosity of the replacement sky - this also helps to eliminate the haloing by reducing the exposure difference between the edges of the foreground and the replacement sky) As a final step, and not always - it depends on the image, I might adjust the blend options for the layer, something like the following. It also helps in reducing the haloing and giving a more natural looking transition. This won't work if you have very strong highlights in the foreground part of the image (as it will then reveal a portion of the replacement image on those strong highlights. The exact shape of the curve would depend on the image.  Needless to say, you should get your image looking right before swapping out the sky, as doing this post swap would likely just accentuate the transition border. 
 
Below is an example image where I used this technique to mask out the sky. First image is a crop from a 400% zoom to show the result, the second is the source image and the last is the final, processed version with the replacement sky added. 

I've tried some of these automated tools and while they work for a good number of images, they don't work all that well on edge cases, which is often when we need them the most. 
There are quite a few different techniques one can use to manually mask out backgrounds depending on what it is you're trying to mask out. 
For difficult daylight sky removals, where there are trees or lots of details on the skyline, I generally use a luminosity mask. This requires multiple steps, so it's not a click and drag, but it gives superior results in my experience. 
In V2 I still do this manually as the new luminosity mask is a live mask and can't be painted on (there are workarounds for using the new luminosity mask, but it ends up no faster than the method below).
Here's my workflow for this particular type of mask
Select --> Tonal Range --> Select shadows Click on Channels panel. Right click on Pixel Selection and choose Create spare channel Now do the same and select midtones instead of shadows In the Channels panel, right click Spare channel and choose Add to pixel selection Now use the use selection brush in subtract mode to deselect areas inside your foreground (and vice-versa if anything in the background is incorrectly selected) Once you have your selection, choose Select --> Grow selection by 1 pixel and then feather selection 4 pixels or less. This will leave a slight halo around your edges if your sky you're masking out is much lighter than the replacement sky, but if you don't do this, then you're likely to get edges that are too sharp and lead to a blocky effect. I never grow by more than one pixel and sometimes feather by less than 4 pixels, never going above 4 (I feathered by 2 on the image below) With the selection still active, click on the mask icon to create a new mask layer from the selection Select the mask layer then alt-click on it to see the black and white mask In sections where the mask should be opaque, paint over it in white if you missed. Ditto for black. Be careful not to go near the edge. We'll deal with that soon. Now click off the layer then reselect the mask layer so you can see the image and not the B&W mask layer Reduce brush opacity to 50% and set brush blend mode to overlay (you can start with even lower opacity and just go over it more times) Zoom in on the border between sky and background and find any areas where there is haloing Now apply the brush (in black) over the edges to reduce the haloing. Make sure you're painting on the mask layer. You may need to increase or reduce brush opacity on the fly. The overlay blend mode is key to getting this step right. You can also apply a curves adjustment to the replacement sky (a U shape, either inverted or not depending on whether you want to reduce or increase the luminosity of the replacement sky - this also helps to eliminate the haloing by reducing the exposure difference between the edges of the foreground and the replacement sky) As a final step, and not always - it depends on the image, I might adjust the blend options for the layer, something like the following. It also helps in reducing the haloing and giving a more natural looking transition. This won't work if you have very strong highlights in the foreground part of the image (as it will then reveal a portion of the replacement image on those strong highlights. The exact shape of the curve would depend on the image.  Needless to say, you should get your image looking right before swapping out the sky, as doing this post swap would likely just accentuate the transition border. 
 
Below is an example image where I used this technique to mask out the sky. First image is a crop from a 400% zoom to show the result, the second is the source image and the last is the final, processed version with the replacement sky added. 

There are a number of tools that don't work in that way. Those would be the ones on the toolbar on the left. The adjustments and filters can work non-destructively on the embedded raw layer. So it's a nice enhancement, but only a partial solution. 
When using a tool that works on a pixel layer, the assistant will auto-rasterize the raw layer. At that point, it destructively applies those develop settings. So you'll have to throw away the edits on the pixel layer if you need to go back and change the develop settings, after which you'll have to rasterize again to a new pixel layer and reapply the edits. 
An example: you output to a raw layer, make a couple of adjustments with say curves and white balance. Since they don't operate on a pixel layer, when you go back into the develop persona and change the develop settings then revert to the pixel persona, the image will reflect both the new develop setting as well as the adjustments you made with the curves and white balance.
But let's say you want to remove a blemish. You select the inpainting brush, copy the raw layer and either rasterize it yourself or let the assistant do it for you. At this point, the develop settings get destructively applied to the pixel layer, which now hides the raw layer. You remove the blemish with the inpainting brush. If you want to go back to alter the develop settings on the underlying raw layer, first you have to hide or delete that rasterized pixel layer (otherwise you'll be changing develop settings on the pixel layer instead of the raw layer). Then you alter the develop settings and go back to pixel persona. You'll see the updated raw layer, but the second you unhide that blemish removal layer, it in turn hides what you did on the raw layer as it's a snapshot of an older raw layer. 
That said, there are a lot of adjustments and filters, so multi-file enhanced-workflow functionality like we're talking about in this thread would be a huge leap forward even if it only works with adjustments and filters and not with the tools that operate on pixel layers only. It'll still need a metadata catalog in order to implement it though, but not necessarily a file format change I think

I can't imagine that especially EIZO provides a 'rubbish' software. I used it without issues for years (but switched meanwhile to X-rite hard-/software to calibrate non-EIZO devices.)
Although I am not familiar with the two .icc versions I agree to @lacerto's hint, and I vaguely remember this also has been mentioned some months ago in an Affinity forums thread as being relevant.
EIZO mentions the difference and possible compatibility issues – although they write v4 would be "backward-compatible" (which sounds conflicting with possible incompatibilty to me).
https://www.eizo.com/library/management/color-management_icc-profiles/

Also this site (linked by EIZO) may interest for a v2 / v4 compatibility test: https://www.color.org/version4html.xalter
To me, opening their PDF in the Apple Preview.app shows incompatibility while opening it in APub V1 as sRGB document looks fine for all four test images / profiles.

Obsolete.

The issue "RAM Usage Limit: Overwriting the numerical values but leaving the MB doesn't keep new values" (REF: AFP-5883) has been fixed by the developers in internal build "2.1.0.1806".
This fix is in the current customer release.
If you still experience this problem once you are using that build version (or later) please reply to this thread including @Serif Info Bot to notify us.

If your photos are too dark when you print them out, it usually means that the brightness on your screen is too high. This results in you getting something that looks "right" on the screen but too dark on the printed photo.
General advice is to calibrate the monitor for 120cmd2 of brightness. I tend to calibrate mine at 100cdm2 as I find even 120cmd2 can result in prints that are a little too dark. 
 
 

No. I'm looking to manually edit the mask in non-contiguous areas. One can unmask areas by painting with the paintbrush but not the reverse. 

Hi @rvst,
Thanks for your post and our sincerest apologies for the delayed response here. We are exceptionally busy following the release of V2 and we thank you for your continued patience and understanding here.
I can confirm I have logged this with our developers as well as a few other issues I have found when using the new Live Mask features.
I hope this helps

This is inadvertently having the wrong effect: I believe the way Affinity colour manages on Windows means that the profile is applied during startup, but cannot be changed or refreshed during app use. Therefore what's happening is that you're disabling colour management entirely, launching Affinity with it disabled, then when you activate the ICC profile that change doesn't refresh within Affinity (until you restart). So the reason Affinity now looks the same as your other software is because nothing is being colour managed.
In your first post, the screenshot comparison with the slight difference in saturation might be because Affinity is colour managing your image correctly. I had a quick search for ImageGlass and colour management and came up with this: https://github.com/d2phap/ImageGlass/issues/43
There seems to be some confusion there between image colour profiles and display colour profiles. The software author is talking about embedded/referenced image profiles and being able to change them, but the key issue here is managing between the image profile and display profile. This is what the Affinity apps do—they will take colour values from the document or image profile (e.g. sRGB, Adobe RGB) and translate them based on the active display profile so that they display correctly when viewed on the monitor.
From reading the above issue on GitHub, it looks like the author implemented the ability to change the image profile, but hasn't implemented the actual translation from image to display profile. Therefore I wouldn't expect ImageGlass to be fully colour managed (only based on that observation above though, please don't just take my word for it).
As far as I'm aware, the Windows desktop composition has nothing to do with the document view in the Affinity apps, so you wouldn't have a situation where colour management is inadvertently applied twice.
It may be worth doing a quick test with an image that uses a wide colour space. I've attached a TIFF of one of my images that's been edited with a ROMM RGB document profile (I've compressed it in a ZIP to prevent the forum software from mangling it!). If you open this TIFF in Affinity and ImageGlass (plus any other software you use), do you notice a big difference in rendering? Affinity should be taking that ROMM RGB profile and translating the colour values based on the custom display profile you have created. Software that isn't colour managed will simply send those colour values to the screen with no translation. You should notice a big difference between a colour-managed and non-colour-managed result with this example.
 
I'm not really sure about this—I use DisplayCAL on both macOS and Windows and wouldn't consider anything else. It sounds like you should maybe just calibrate and profile with DisplayCAL, use its own calibration loader and then assume what you see in Affinity is correct. You can experiment with other apps as well, but do check that they perform document-to-display (or image-to-display) colour management, and don't just offer an option to override the image profile being used.
 
A useful diagnostic "hack" you can use within Affinity Photo is to go to Document>Assign ICC Profile (not Convert) and assign your display profile to the document—for example, mine might be UP3216Q #1 2022-09-19 11-11 D6500 2.2 F-S 1xCurve+MTX. This will effectively bypass colour management and show you what the image would look like if its colour values weren't being translated. If your document is only in sRGB, you might notice a very minimal change, if anything at all—you might see a small shift in saturation like with your first landscape image example. If it's in a wider space such as Adobe RGB or ROMM RGB, however, you should see a more noticeable difference.
And on that note, a general rule to observe: always use device or standardised profiles for your document (sRGB, Adobe RGB etc), never display profiles. Display profiles should only be used by the OS and software to colour manage between the image/document and display.
Hope the above helps in some way!
JR ROMM RGB_6030007 8-bit.tiff.zip

Hi, this topic came up recently.
 

Run the 2.0.4 installer and the 2.0.3 will install cleanly over, so 2.0.4 will install to the same folder, without asking (it will update)

There's a completely reworked and much cleaner variant of this code on my github, although to be fair, I'd suggest anyone just use the test 2.0.3 MSI installer instead. No reason to use the MSIX anymore. 

We are not yet finished with all our testing of these but they have passed initial testing and are ready for use to outside testers. 
These UNSANDBOXED (MSI/EXE) installers are available to use, hosted on Dropbox while we complete our testing.  IF you would like to try these please do so, but as I say they are currently unofficial.
Affinity Designer 2.0.3.1688
Affinity Photo 2.0.3.1688
Affinity Publisher 2.0.3.1688
I think that these will soon be replaced with official 2.0.4 ones but you are welcome to use these and @ me with what you find.
You can simultaneously have BOTH the Sandboxed (MSIX) and the Unsandboxed (MSI/EXE) installed, but note they are hard to tell apart.
You cannot run a mixture of both MSIX and MSI, at the same time that will NOT work.

I should probably give DxO a look again. I last bought a copy around a decade ago and it was a little bit of a one-trick pony at the time. Based on your comment and the length of time since I last had a copy, it might be a worth a spin

So your GPU trounces mine on the Geekbench OpenCL benchmark, a full 44% faster, but is 30% slower on the Affinity benchmark.
Puzzling. 

Why? The benchmark calculation is for our diagnostic purposes, we can change it when we wish.

We can't really say that. The benchmark across different versions is not directly comparable, even minor versions.

When I tested on slower hardware combinations, the gap seemed lesser (%-wise). Which was strange. But his would be margin of error on my main rig.