Please check this out. I started by using the blue channel layer and worked with to create a mask. https://drive.google.com/file/d/1OZykpp ... sp=sharing
Isolate hair
-
- Mitglied
- Posts: 216
- Joined: Sat 28 Sep 2024 17:04
Re: Isolate hair
-
- Mitglied
- Posts: 120
- Joined: Tue 18 Aug 2020 03:07
Re: Isolate hair
Asked me to send request to download.Vivi_Ram wrote: ↑Fri 17 Jan 2025 19:07 Please check this out. I started by using the blue channel layer and worked with to create a mask. https://drive.google.com/file/d/1OZykpp ... sp=sharing
Thanks
-
- Mitglied
- Posts: 216
- Joined: Sat 28 Sep 2024 17:04
Re: Isolate hair
@lacogada. Don't know if you requested the access to that file. I've granted it. If you send me link to that image you're working on, I can try and build a mask to demonstrate my approach to it. I'll record the video for you.
-
- Mitglied
- Posts: 120
- Joined: Tue 18 Aug 2020 03:07
Re: Isolate hair
Wow that's a very good cut out ... thanks. Sent PM
-
- Mitglied
- Posts: 120
- Joined: Tue 18 Aug 2020 03:07
Re: Isolate hair
-
- Mitglied
- Posts: 120
- Joined: Tue 18 Aug 2020 03:07
Re: Isolate hair
-
- Mitglied
- Posts: 2353
- Joined: Sat 12 May 2012 21:38
Re: Isolate hair
Here's the example.

PLD file: https://gofile.io/d/GzM13f
While I was adjusting the channels, I quickly noticed that the blue of the sky mixed in a lot with the bridge's cables. And removing it like that guy does in the video left in fringe edges all over the place (which his suffers from as well). Crunching the blue channel too much will result in thinning the cables to the point that parts of the aliased cable chunks merely disappear - which is unwanted.
I could only prevent that by mixing in parts of red and green channel and fine-tune the mask parts. That's where a Photoshop user would have to keep undoing and trial-and-erroring it destructively, while I could just non-destructively adjust the settings and see the mask result in real time.
That said, with darker background replacements mine still has obvious fringing, and that would have to be fixed manually using the same methods that are explained in the video.
Otherwise is pretty much the same approach as that video. The main difference is that this PhotoLine version is entirely non-destructive but for the filler mask layer which takes care of manual mask edits.
A second important difference: I reuse the final mask as a cloned version, simply invert that clone with an adjustment layer, and then adjust the colour temperature of the foreground city to better match the new background.
Anyway, you can tell from watching the Photoshop tutorial that they are copying and pasting masks and mask parts destructively, so it just is impossible to backtrack and adjust the original mask. It takes so much more time to experiment in Photoshop because of its layer mask and layer stack limitations. With this file anyone here can have a go at improving the mask themselves by adjusting the adjustment layers. The clones automatically update.

PS I did encounter layer thumbnail corruption at times when I worked with the layer property channel checkboxes. And that also affected the screen when panning the view: it caused temporary screen artefacting.
PS2 I checked the file again, and in my rush to create this example I admit it's rather rough around the edges. Could be cleaned up in various places, of course. Hopefully it will motivate others here to experiment around.
/*---------------------------------------------*/
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
-
- Betatester
- Posts: 4097
- Joined: Sun 03 Jul 2005 13:35
- Location: Deutschland
Re: Isolate hair
I agree a lot with Herbert123, and nice that we meet for experiments again 
My thoughts about the video:
- It is REALLY nice that he put a link to the used picture in his video description! Liked him just for that.
- "Calculations" to combine channels has some options that the channel mixer hasn't. He's using the tool with trial&error. Not looking at the picture and channels, deciding what to do and finally make the right settings.
- - He noticed that only a few calculations are really useful. Adding and subtracting can better be done in the channel mixer. Multiplying not so much.
- - Having Lab and CMYK channels easily available for trial%error is nice. But CMYK is just inverted RGB with some even unwanted adjustments for K, Lab is (for this purpose) not better than the very similar YCrCb. Channel mixer can handle that - if you know how to set the numbers...
- - In the end he added blue to blue with some offset. Selecting blue and multiplying by 2 can be done with curves or levels, he needs these tools anyways.
- Dodge and burn on the mask... The sky had a gradient, so the mask isn't perfect everywhere. Dodge&burn is a quick hack to make some local changes to the mask, looks good enough here. For more complicated, semi-transparent stuff like hair, painting on a curves adjustment would be better.
- I'd advice to keep the natural order of layers! Background is background, and foreground is foreground. So the sky should be the simple background layer, and the bridge should be selected and put above that, not the other way around. Makes no difference in such simple example, but in more complicated composings (like placing a moon behind the bridge) I had the "shit, should have followed the rule" too often.
- I would handle these fine fringes just like that, by painting. Sometimes a special mask that only selects the fringes can be useful, and then smudge or minimum/maximum can be nice, too.
- Using "blend if" as a quick hack because painting the fringes destroyed details of the bridge... Works good enough here. He's aware that he wants to restore the "original" color of the bridge, and that he could introduce fringes again.

My thoughts about the video:
- It is REALLY nice that he put a link to the used picture in his video description! Liked him just for that.
- "Calculations" to combine channels has some options that the channel mixer hasn't. He's using the tool with trial&error. Not looking at the picture and channels, deciding what to do and finally make the right settings.
- - He noticed that only a few calculations are really useful. Adding and subtracting can better be done in the channel mixer. Multiplying not so much.
- - Having Lab and CMYK channels easily available for trial%error is nice. But CMYK is just inverted RGB with some even unwanted adjustments for K, Lab is (for this purpose) not better than the very similar YCrCb. Channel mixer can handle that - if you know how to set the numbers...
- - In the end he added blue to blue with some offset. Selecting blue and multiplying by 2 can be done with curves or levels, he needs these tools anyways.
- Dodge and burn on the mask... The sky had a gradient, so the mask isn't perfect everywhere. Dodge&burn is a quick hack to make some local changes to the mask, looks good enough here. For more complicated, semi-transparent stuff like hair, painting on a curves adjustment would be better.
- I'd advice to keep the natural order of layers! Background is background, and foreground is foreground. So the sky should be the simple background layer, and the bridge should be selected and put above that, not the other way around. Makes no difference in such simple example, but in more complicated composings (like placing a moon behind the bridge) I had the "shit, should have followed the rule" too often.
- I would handle these fine fringes just like that, by painting. Sometimes a special mask that only selects the fringes can be useful, and then smudge or minimum/maximum can be nice, too.
- Using "blend if" as a quick hack because painting the fringes destroyed details of the bridge... Works good enough here. He's aware that he wants to restore the "original" color of the bridge, and that he could introduce fringes again.
Yes, that mixing of channels is even quite predictable. Pick a typical red-orange-dark of the bridge, substract a typical light blue of the sky, put the result into the channel mixer, play with factors and offset.Herbert123 wrote: ↑Sat 18 Jan 2025 11:10... I could only prevent that by mixing in parts of red and green channel and fine-tune the mask parts...
Wouldn't do that, a mask and the inverted version don't combine to 100% for adjustments. Just apply the color temperature to the whole picture of the bridge.Herbert123 wrote: ↑Sat 18 Jan 2025 11:10A second important difference: I reuse the final mask as a cloned version, simply invert that clone with an adjustment layer, and then adjust the colour temperature of the foreground city to better match the new background.
----------------
Herr Doktor, ich bin mir ganz sicher, ich habe Atom! /Doctor, doctor, I'm sure, I've got atoms!
Herr Doktor, ich bin mir ganz sicher, ich habe Atom! /Doctor, doctor, I'm sure, I've got atoms!
-
- Betatester
- Posts: 4097
- Joined: Sun 03 Jul 2005 13:35
- Location: Deutschland
Re: Isolate hair
Later more, just quick: The color cube plugin is 32bit only, therefore it won't work in PL anymore. Does someone know a working replacement? That tool helps a lot with "knowing what to do". It's like a 3D histogram, showing the used colors in a 3D cube.
You do not have the required permissions to view the files attached to this post.
----------------
Herr Doktor, ich bin mir ganz sicher, ich habe Atom! /Doctor, doctor, I'm sure, I've got atoms!
Herr Doktor, ich bin mir ganz sicher, ich habe Atom! /Doctor, doctor, I'm sure, I've got atoms!
-
- Mitglied
- Posts: 216
- Joined: Sat 28 Sep 2024 17:04
Re: Isolate hair
[/quote]
Here's the example.

PLD file: https://gofile.io/d/GzM13f
[/quote]
Brilliant, thank you very much @Herbert123. I'm going take some time to look at the file carefully and update this comment with more details. Watch this space. Appreciate you posting the file!
Here's the example.

PLD file: https://gofile.io/d/GzM13f
[/quote]
Brilliant, thank you very much @Herbert123. I'm going take some time to look at the file carefully and update this comment with more details. Watch this space. Appreciate you posting the file!
-
- Mitglied
- Posts: 2353
- Joined: Sat 12 May 2012 21:38
Re: Isolate hair
I've setup the 32bit version of PhotoLine as an external program plugin inside PhotoLine 64bitHoogo wrote: ↑Sat 18 Jan 2025 15:05 Later more, just quick: The color cube plugin is 32bit only, therefore it won't work in PL anymore. Does someone know a working replacement? That tool helps a lot with "knowing what to do". It's like a 3D histogram, showing the used colors in a 3D cube.
So I can access all the 32bit plugins easily in another instance of PhotoLine (the 32bit version) that way, and send the result back to PhotoLine 64bit

...and thanks for the colorcube plugin link. I wasn't aware of that plugin. That is actually quite handy!
You do not have the required permissions to view the files attached to this post.
/*---------------------------------------------*/
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
-
- Mitglied
- Posts: 2353
- Joined: Sat 12 May 2012 21:38
Re: Isolate hair
Agreed!

Often these tutorial youtubers fail to link to the file(s) that they use in their videos, and it is nice that he includes it.Hoogo wrote: ↑Sat 18 Jan 2025 15:02 My thoughts about the video:
- It is REALLY nice that he put a link to the used picture in his video description! Liked him just for that.
- "Calculations" to combine channels has some options that the channel mixer hasn't. He's using the tool with trial&error. Not looking at the picture and channels, deciding what to do and finally make the right settings.
- - He noticed that only a few calculations are really useful. Adding and subtracting can better be done in the channel mixer. Multiplying not so much.
- - Having Lab and CMYK channels easily available for trial%error is nice. But CMYK is just inverted RGB with some even unwanted adjustments for K, Lab is (for this purpose) not better than the very similar YCrCb. Channel mixer can handle that - if you know how to set the numbers...
- - In the end he added blue to blue with some offset. Selecting blue and multiplying by 2 can be done with curves or levels, he needs these tools anyways.
- Dodge and burn on the mask... The sky had a gradient, so the mask isn't perfect everywhere. Dodge&burn is a quick hack to make some local changes to the mask, looks good enough here. For more complicated, semi-transparent stuff like hair, painting on a curves adjustment would be better.
I also had the distinct impression that he was sort-of 'feeling his way around', rather than taking a more strategic approach. While reproducing his steps, I found I just got way too much fringing, and he gets around this by manually cleaning things up near the end.
But on the other hand it is great to see actual channel-based selections making a comeback and showing how the AI generated mask just isn't as good.
The channel mixer worked well to reduce the fringing and keep the cable details well enough to work for this particular background.
Totally agree! Usually I order my comps according to that rule myself, and I had the same thought while replicating the steps, but left it as the original tutorial.Hoogo wrote: ↑Sat 18 Jan 2025 15:02 - I'd advice to keep the natural order of layers! Background is background, and foreground is foreground. So the sky should be the simple background layer, and the bridge should be selected and put above that, not the other way around. Makes no difference in such simple example, but in more complicated composings (like placing a moon behind the bridge) I had the "shit, should have followed the rule" too often.
Yes, I had to resist the urge to clean up these fringes manually.

A quick maximum (round) adjustment layer destroyed the cable details too much, though. Even at very subtle settings. The cables in this photo are already very anti-aliased and mixed with the background sky a tad too much. I am guessing this photo was reduced in Photoshop, and my experience tells me that CatmullRom retains contrast in details like these better than Photoshop's downsampling.
Tried it, did not like the effect it had on the details. Or at least, it did not achieve what I thought it would achieve? Could be my mistake, though.
I actually quite like (in this particular example) how I can separately control both the sky's and foreground's warmth here. Gives it more of a 'filmic' cinematic look in my opinionHoogo wrote: ↑Sat 18 Jan 2025 15:02Wouldn't do that, a mask and the inverted version don't combine to 100% for adjustments. Just apply the color temperature to the whole picture of the bridge.Herbert123 wrote: ↑Sat 18 Jan 2025 11:10A second important difference: I reuse the final mask as a cloned version, simply invert that clone with an adjustment layer, and then adjust the colour temperature of the foreground city to better match the new background.

/*---------------------------------------------*/
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
-
- Mitglied
- Posts: 1772
- Joined: Mon 23 Dec 2019 15:21
- Location: Ukraine
Re: Isolate hair
Fringing artifacts could be due low resolution and too sharpened source photo. Some wires on that bridge literally have 1 pixel thickness which is just not enough to create quality mask.
Really interesting tutorial for both Ps and PL anyway!
Really interesting tutorial for both Ps and PL anyway!
PhotoLine UI Icons Customization Project: https://www.pl32.com/forum3/viewtopic.php?f=3&t=6302
-
- Mitglied
- Posts: 2353
- Joined: Sat 12 May 2012 21:38
Re: Isolate hair
Yes, agreed. It's obviously a downscaled photo, and I encountered jpeg compression artefacting in the subtle sky gradient while generating the core mask!
And the particular resampling method used while downsampling image data can have a very obvious effect as well.
/*---------------------------------------------*/
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
-
- Mitglied
- Posts: 2353
- Joined: Sat 12 May 2012 21:38
Re: Isolate hair
By the way, I think I found a bug. Can other test this:
[1] Open the file that I linked to earlier: https://gofile.io/d/GzM13f
[2] Navigate to the Channels virtual layer.
[3] In the Layer properties turn OFF the R and/or G channel(s).
Result: a few layer thumbnails are corrupted: the Channel Mask layer, for example. And the Final mask.
Zoom in and out into the layer stack (<CTRL> scrollwheel) to experience funky looking "layer effects"!
p.s.: while working on this file with channels activated and deactivated here and there in the layer stack I also experienced that the view would sometimes get corrupted while panning it (the areas hidden in the view would scroll into view with line artefacting).
[1] Open the file that I linked to earlier: https://gofile.io/d/GzM13f
[2] Navigate to the Channels virtual layer.
[3] In the Layer properties turn OFF the R and/or G channel(s).
Result: a few layer thumbnails are corrupted: the Channel Mask layer, for example. And the Final mask.
Zoom in and out into the layer stack (<CTRL> scrollwheel) to experience funky looking "layer effects"!

p.s.: while working on this file with channels activated and deactivated here and there in the layer stack I also experienced that the view would sometimes get corrupted while panning it (the areas hidden in the view would scroll into view with line artefacting).
You do not have the required permissions to view the files attached to this post.
/*---------------------------------------------*/
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait
System: Win10 64bit - i7 920@3.6Ghz, p6t Deluxe v1, 48gb (6x8gb RipjawsX), Nvidia GTX1080 8GB, Revodrive X2 240gb, e-mu 1820, 2XSamsung SA850 (2560*1440) and 1XHP2408H 1920*1200 portrait