As Mr.Cox has explained (might have been in another thread) Adjustment Layers can work fast because the calculate their effect for each pixel based on that pixel’s values alone, Filters on the other hand can derive the resulting values for a pixel from a huge number of (other) pixels, so the task is much more complex and does not lend itself to a Filter Layer implementation easily.
I think this would be very powerful and helpful even if there's a "slow" responses, and even if there needs to be a "render" button or sorts.
Just like when you try to resize a smart object with smart filters and it warns you that the smart filters won't display until the transformation is done, the same could apply when you move a layer under the adjustment layer.
We actually tried the idea at one time.. and prerelease users used a lot of expletives and creative invective to describe how slow and unwieldy it was.
It's not just slow, it's too EXPLITIVE slow to use.
Adjustment layers are one of the most powerful tools in Photoshop. I love them, but they have a pretty limited selection. As a game developer I have often come across the need for the ability to write my own to accomplish my goals more efficiently - specifically with the ability to see the results of the filter in real-time like adjustment layers. Use cases include visualizing/authoring texture channel data for shaders, or a real-time indexed palette filter (as described in my blog http://danfessler.com/blog/hd-index-painting-in-photoshop)
My suggestion is to create an API to create custom adjustment layers. It turns out Adobe has already done quite a bit of the legwork with pixel-bender. If there were a new adjustment layer type that accepted pixel bender files, then this would be a perfect solution (and is GPU accelerated to boot)
Hey Chris, thanks for the response in the other thread.
"1) No, you can't have filters as adjustment layers, the math just doesn't work in the user's favor (i.e.: too slow to be useful)"
I understand that filters would be too slow, which is why I didn't request filters (also why it doesn't make sense to merge my topic with this one, I read up on previous posts regarding smart filter requests). This feature would need to be it's own API independent of filters that operate purely on individual pixels within a single pass given provided inputs. This would have the same performance footprint as any other adjustment layer.
"2) PixelBender is gone and will not be coming back."
The former is why I suggested PixelBender specifically - it meets the performance criteria and can be limited to single-passes. I'm not sure of the reasons why pixel bender died, it sounds unfortunate, but the premise of my suggestion still stands. If not Pixel Bender, then simply a dedicated API that has limitations which favor performance would suffice.
Would this approach to the problem resolve any technical resistance to adding such a feature?
It is basically the same request. No, PixelBender is still a filtering operation and cannot be used like an adjustment layer - the performance would be horrid.
An adjustment layer API might be possible, but not easy. We've shown some third party developers the limitations we have to live with to keep high performance and stability - and they had a lot of trouble trying to understand and deal with those limitations.
You would definitely know more about pixel bender's performance limitations than I would but my understanding of it was it was basically a fragment shader language whose very purpose is generally for high performance real-time redraw via the GPU which I'm quite familiar with (hello from the Zynga building from across the street). As with any tool though, you of course can abuse it past the point of usefulness if you're careless. This even includes the existing tools in photoshop of course. But discussion about pixel bender is moot I suppose because, as you've mentioned, it's dead.
In any case, I would absolutely LOVE to see a dedicated API for this. I can think of many applications - even with the limitations. Every year or so I check to see if one exists and am disappointed.
Solution: applying filters to folders or as in adjustment layers Smart objects is slow. You need to be grouping into smart objects to apply the same effect. Adobe we need more productivity. This solution would be great.
This program uses smart objects directly without having to convert each image. Also uses non-destructive layer filters that affect more than one layer. He also has a layer mask for each filter.
Yup, PhotoLine has also been applying live filters to layers without the need for smart objects for years and years now.
However, Chris Cox is correct in stating that at a certain point of complexity live filters do slow down things. At that point smart objects can alleviate performance bottlenecks in your layer stack. Affinity Photo does not yet support smart objects. PhotoLine does, which gives it the best of both worlds.
37
Replies
AdChoices
