P: Improving the Graduated Filter

Currently just CC, but you can get an early glimpse what's coming to the raw pipeline. 🙂

Lightroom too, soon? or just Photoshop CC...

Yes, totally agree. This would be an excellent improvement to Lr.

I'd love to have the ability to shape the Graduated Filter. Lots of times when I apply it to a sky in order to pull down highlights, introduce more saturation and/or add contrast to moody clouds etc, it just doesn't live up to it's potential because I'm hindered by it being a straight line. If there's a tree/hill/mountain/ or whatever on the horizon, then my ability to achieve what I want is compromised.

I totally agree with Sean on this; it's all about working efficiently and quickly. The graduated filter is so powerful, in an often quite subtle but effective way - I use it all the time just to shift the emphasis in an image a little. Being able to warp it, in an easy-to-use and straightforward way, around foreground objects or into valleys, would be excellent.

Patrick Cunningham

I like the idea of allowing the outer shape (not the graduation) of the graduated filter to be controlled with a polygon with a variable number of points. Seems more useful than tying brush strokes to a graduated filter.

Yeah, I'm not opposed to Sean McCormack's idea either - but if I had to choose...

There's a reason why people are gah-gah over U-points, and why Nik Software can charge $200 for their plugins (hint: its not because their adjustment algorithms are *so* good, or their revolutionary UI).

There's also a reason why Adobe chose an auto-masking brush instead of developing other auto-masking technologies. I don't know what those reasons are, and I don't know what the limiting factors are in Lightroom that might make it prohibitive when it has not been prohibitive in other parametric / non-destructive environments.

But in the case of sky above mountains:

Estimated user time for selecting sky using U-points: 2 seconds (1 U-point in the sky, and optionally 1 anti-U-point on the mountain).
(If there were tricky sky-like elements to be excluded, which require additional attention, it could take several seconds. For the most complex cases, it could even take a minute or two...)
Estimated data storage: 6 words: position, radius, and opacity for each U-point (for typical case of 2 U-points), maybe several dozen bytes for more complex cases.
Estimated time to compute mask, as tested in NX2: 0.01 seconds (it appears to be instantaneous).
Estimated time to render adjustments tied to mask, as tested in NX2: 0.1 seconds (some adjustments appear to be instantaneous, others take a small fraction of a second - total time depends on the number and types of adjustments tied to the mask). Once rendered, they are never re-rendered again during development unless one changes adjustments that precede it in "the stack".

Summary:
--------
Auto-masking selection technology in a parametric/non-destructive environment is doable and can be efficient - we have proof.
Whether there is something about the design of Lightroom that makes it especially ill-suited for such technology - I don't know.

Final Thoughts:
---------------
- U-points have their limitations too, which is why I'd like to see Adobe do something different and better.

I'm not opposed to Sean's idea, and if Adobe does not invent something better, or maybe even if they do, then I hope it goes through...

I think it should be our job to suggest the features we want, and then leave it for Adobe to figure it if they want to support and if it's technically feasible. I never would have imagined that they could make the brush tool work as well as it does, so why would I presuppose that I can decide whether they can make this suggestion work well too? We're supposed to be the creative ones and creativity is often rewarded with amazing implementation!

I'd like Lightroom to be usable. We're already on the threshold of what's doable as a metadata edit. Mixing complex masks, lens corrections and spotting, and then trying to do further edits is pushing the limit as to what it can happen.

There is a time when a rendered file is the way to go and that is the time. If it's an issue, then use Smart Objects, but I want usable tools for 90% of stuff. The tree on the horizon? In real life I could care less about how dark it goes. I personally wouldn't bother masking it. Following a ridge line on the other hand is something I would like to do.