Please refer the short Video tutorial talks about the graduated filter in Lightroom
We have tried using the sharpening effect at our end and yes it does make changes once you past over -50 in sharpening effect slider.
Thanks for that - I was just expecting that it would be more gradual between the two lines, not such a sudden visible line.
I've been having this same problem since a few months ago. I used to work gradually for me for the past 5 years! And then, all of a sudden, it stopped working. It seems to be a bug. It always applied the blur gradually from 0 to a 100, but now, no matter what I do, I get an abrupt line where it's zero blur on one side and 100% blur on the other!
I really don't know what happened, but it is very annoying and unusable now.
no matter what I do, I get an abrupt line where it's zero blur on one side and 100% blur on the other!
I just tried a quick test, applying a graduated filter with Sharpness = 100 and no global sharpening in my CC 2015.12. I observed a smooth gradation:
Do Help > System Info and report the exact version of LR you're running.
Select a sample photo showing the problem (i.e. with a graduated filter applied that's not smoothly blurring). Do Metadata > Save Metadata To File, upload the file and its .xmp sidecar to Dropbox or similar, and post the sharing link here. We can see if the issue happens in other LR configurations.
This is what I get in Help > System Info:
Lightroom version: CC 2015.12 [ 1125239 ]
License: Creative Cloud
Operating system: Mac OS 10
Version: 10.11 
Application architecture: x64
Logical processor count: 8
Processor speed: 2.6 GHz
Built-in memory: 16384.0 MB
Real memory available to Lightroom: 16384.0 MB
Real memory used by Lightroom: 2227.8 MB (13.5%)
Virtual memory used by Lightroom: 5747.8 MB
Memory cache size: 1745.4MB / 3864.7MB (45.2%)
Maximum thread count used by Camera Raw: 5
Camera Raw SIMD optimization: SSE2,AVX,AVX2
Camera Raw virtual memory: 1634MB / 8191MB (19%)
Displays: 1) 2880x1800
Graphics Processor Info:
NVIDIA GeForce GT 750M OpenGL Engine
Check OpenGL support: Passed
Vendor: NVIDIA Corporation
Version: 4.1 NVIDIA-10.10.14 310.42.25f02
Renderer: NVIDIA GeForce GT 750M OpenGL Engine
Here is the Dropbox link to a sample file and its .xmp sidecar experiencing this issue:
1 person found this helpful
[This post is best viewed on the forum in your browser, not email, due to the numerous embedded images.]
This was a curious puzzle, so I investigated a bit. What I've learned is that the graduated filter always varies sharpness values smoothly from no blurring to maximum blurring. But the narrower the width of the filter, the more likely it is that our eyes will perceive even a small amount of blurring of fine details as a sharp boundary.
Stacking multiple instances of the filter effectively narrows the width of the transition from no blurring to full blurring. And the narrower the width of transition, the more likely it is that our eye will perceive the start of even the smallest amount of blurring as a distinct boundary. Your sample photo stacked 8 instances of the filter, effectively cutting the width of the transition to 1/8 of the filters' width, producing a very visible boundary where the blurring started.
- There's no point in stacking multiple instances of the filter -- you get the same results just by narrowing the width of a single filter.
- If you use the maximum sharpness value of -100, you need to use a wide graduated filter, otherwise our eyes will perceive the smooth varying of values as a sharp boundary.
It used to work gradually for me for the past 5 years!
I tested with both LR CC 2015.12 and LR 5.7.1, observing the same results in both process version 2010 and 2012. So if there has been a change in behavior, it must have occurred at least several years ago, prior to LR 5.7.1. Or perhaps you changed your method somewhere along the way?
As previously observed, sharpness values of the graduated filter between 0 and -50 subtract from the global sharpness value but don't add blur. Values between -50 and -100 add increasing amounts of blur.
The blur increases smoothly as you shift the slider from -50 to -100, as you can see in this video. The blur goes from just a very small amount on fine details (-50) to a very large amount on large, coarse objects (-100).
When you look inside the filter boundaries, the right edge is 0, the center line is -50, and the left edge is -100. So you won't see any blurring between the right edge and the center line (0 to -50) -- the blurring starts to the left of the center line and gets stronger towards the left edge.
When the left and right edges are far apart, spanning most of the photo, the blur increases smoothly from left of the center to the left edge:
The finest details start blurring fairly quickly but smoothly as you move left from the center.
When you shrink the distances between the left and right edges, the transition from the center to the left edge must necessarily occur more quickly, in a shorter width. The transition from non-blurring to blurring is more noticeable but still smooth:
But if you make the distance between the edges even smaller, the transition where the blurring starts starts looking like a distinct boundary:
This is because the blur has to transition from no blur (-50), to a tiny bit of blur on fine details (-51), to full-on blur of large coarse objects (-100) in a very short distance. Even though the value of the blur increases smoothly, the fine details start blurring heavily in a very short distance, and that's what our eye perceives as a distinct boundary.
In your sample photo, you had 8 graduated filters stacked on top of each other, each with its slider set at -100. This effectively narrows the width of the transition from -50 (smallest blur) to -100 (largest blur) by a factor 8! This ensures that the transition from no blur to a small amount of blur happens in a very short distance, and our eye perceives that transition as a sharper boundary.
It's a little counter-intuitive why this happens with stacking. To understand, first let's duplicate the filter from the previous example (right-click on the pin and select Duplicate), so there are just two instances of the filter Sharpness = -100 stacked on top of each other:
Notice that the boundary appears much more distinct, and that it has shifted right. Why?
When filters are stacked, LR adds the values of their sliders (* footnote). So with one instance of the filter, the sharpness value goes from 0 on the right edge, to -50 in the center, to -100 on the left. This is represented by the orange line in the graph below:
The blurring only occurs for values between -50 and -100, between the center and the left edge.
When you place a second instance of the filter, LR adds their sharpness values, yielding 0 on the right edge to -200 on the left edge, as shown by the green line.
But LR caps sharpness values at -100, so the actual result of adding the two filters is shown by the blue line: 0 at the right edge, -100 at the center, and -100 at the left edge. The blurring occurs for values between -50 and -100, which extends from halfway between the right edge and center to the center.
This shows graphically that with two filters, the width of the transition from no blurring to maximum blurring is 1/2 the width of one filter. When you stack eight filters, the width of the transition will be much narrower, 1/8 of the width of one filter.
And as shown above, the narrower the width of the transition, the more likely it is that it will be perceived by eye as a sharp boundary.
To verify my conclusions, I did more precise experiments using a fine checkerboard pattern and comparing before and after results with Photoshop layers. This let me determine exactly where even small amounts of blurring started and where increases in blurring stopped.
(* footnote) Experiments show that LR doesn't strictly add the sharpness values of stacked filters -- subsequent filters get a little less weight. But the principle and overall effect is the same.
Thank you for looking into the issue in such detail johnrellis!
I'm just not sure why Adobe has designed the gradation tool on sharpness in such a way that it only starts blurring or sharpening in the middle and not evenly from start to finish. And the reason I stack several on top of each other is that I can't go below -100 and the blurriest section is not blurry enough for me. Maybe you have a workaround for that?
I may have been mistaken, and this issue was happening before too and I just recently noticed it.
I guess for now I'll stick with Lens Blur in Photoshop which works like it should and results in flawless graduated blurs, just more work and a bit of an inconvenience.
I'm just not sure why Adobe has designed the gradation tool on sharpness in such a way that it only starts blurring or sharpening in the middle and not evenly from start to finish.
It is a little counter-intuitive. I think they wanted the filter sharpness slider to be able to completely undo any global sharpness values (just like the filter exposure slider can undo global exposure values), and they wanted the slider to be able to add blur as well. Offhand I can't think of another way of accomplishing both of that with a better design than the current one, but it's likely someone else could come up with a better idea.
And the reason I stack several on top of each other is that I can't go below -100 and the blurriest section is not blurry enough for me. Maybe you have a workaround for that?
No workaround, other than going into Photoshop. Each develop adjustment has a maximum value, and the filters just add their slider values before applying them. So there's just no way to get more than sharpness = -100, unfortunately.