Finally ditching Photoshop

Joanna, I’m finding on Mac that the new clone tool is very hit and miss and very slow to correct. The mechanism of choosing the clone origin for each clone action is brilliant but if it’s this slow, then for more than a very small correction, it’s just too slow.

I’m not finding that I can get good results very fast, even on fairly easy removals like a bag on grass. The area of correction which you showed (white water) has a fairly high tolerance for error but even by those standards the area cleaned up looks tampered with.

As great as the clone tool concept is, the actual tool urgently needs some fine tuning and a radical speed boost. Serious cloning is difficult in a RAW editor as theoretically any changes to tones should pass through to the cloned area resulting many additional interdependent calculations for each layer of the clone. Perhaps for the sake of speed, cloning should be left to last and become a single bitmap layer of local adjustment rather than remaining part of the RAW original. Coming back to edit the image would then result in having to redo the cloned area. Or perhaps new global adjustments could be applied but not to RAW data but to bitmap data.

A viable clone tool for RAW editing is extremely difficult to engineer. Lightroom’s tool in version 4 was no great shakes. There is no RAW clone tool in Photoshop’s ACR. Clone/repair is only done on the bitmap image.

Perhaps you have some tips on how to make the clone tool run faster or how to work more effectively with it?

Alec,

With regard to the application of cloning masks I found a solution that significantly sped up the process for me. I’m on Windows 10 rather than a Mac, as I believe you know, so this may not be related to the slowness you are seeing. I found the cloning performance very significantly slower on some images than on others. There didn’t seem to be any particular reason for this and I was living with it. I started to delve deeper when I became frustrated trying to clone out a number of objects on a particular image that was taking forever.

Then, as a result of the new local adjustment palette, I noticed it. I had four local adjustment masks on that image. Playing around with other images with no local adjustments I found cloning was significantly faster. I created a virtual copy of the image with the four masks and tried cloning again with the same slow results. Then I deleted all the local adjustments from that virtual copy and tried cloning again . Applying each cloning mask again took a second or two at most and was occasionally almost instantaneous. I then reapplied the local adjustments which did not appear to take longer as a result of cloning first. Going forward I plan to apply repair and cloning edits before adding local adjustments.

Of course as I said the situation on your Mac may be caused by something completely different. I also have no idea whether or not you had applied local adjustments prior to cloning, but I thought I would share what I found.

Mark

Edit: It also appears that hiding all the LA masks AND deselecting the local adjustment palette has a similar effect on cloning performance. In a quick test it seems that both hiding masks and deselecting the palette has a greater effect on cloning performance than just doing one or the other, but I need to do more testing.

If others can confirm that local adjustments do affect cloning performance, at least on Windows, as my tests suggest, then this needs to be documented for users and addressed as a performance issue by DXO.

3 Likes

Actually there is and it’s been there for some time… “Spot Removal” which has two modes: “Clone” & “Heal”. You can set-up a group of images to use the same area to repair and source area for repair - which is great if You have sensor/lens dust in a bracketing sequence.

And it works extremely fast.

The set-up:
Repair%20ACR

The settings for the tool:
Repair%20Clone%20settings

Indeed. Great demonstration.
We should be partners Peter for a DxO school :stuck_out_tongue_winking_eye:
Pascal

2 Likes

Volontiers.
Tu clones une première zone, l’image est modifiée.
Tu peux utiliser ce premier résultat comme source d’une nouvelle correction.
adobe ne sait pas le faire dans un workflow raw :wink:
Pascal

You clone a first zone, the image is modified.
You can use this first result as source of a new correction.
adobe can not do it in a raw workflow :wink:

1 Like

One thing which is now a interesting feature, you can change from repair to clone and back on excisting editpoints.
Just select a pin and change settings.

Other thing repair is more a onestroke mask thing wile cloning can be used as knibbeling feature to replace a group of pixels.
But you can knibbel with repair and change halfway to clone by reselect pins and change the function.

Other thing: the not seen image by the mask as a plain blue and red blob is in backlog as improvement request.
As in speed, does zooming in helps? Less preview rendering i suppose does speedup the preview of changed pixels.

1 Like

Thanks for these tips Mark. I’ll play around with them and see if I can’t get the clone/repair tools to work at a usable and useful pace. PhotoLab should automatically disable display options which slow down the clone/repair tool when it’s enabled. These are new features so DxO is no doubt working on optimising them. Field reports like ours are what they need to know where the pain points are though.

Taking into account all your comments and hints, I’ve just reworked the image again.

Difficult to tell if zooming helps because I am usually quite zoomed in already to do the retouching.

I didn’t notice that much difference in hiding masks and disabling the palette, at least for what I was doing.

Pascal, I now understand what you meant about using already cloned areas as a source - thank you, it was very useful in my new test.

I will add another comment on using the clone tool, however. Having to hide the masks to see the effect of the cloning, then having to reshow them to make the next correction, is a real nuisance. I feel having a dotted coloured outline would be far better than a filled area - at least whilst editing an existing cloned region. I seem to be struggling for how to describe what would be best - can someone please help argue this out?

1 Like

Put the mouse cursor out of image window.
Pascal

1 Like

As of now, DPL’s implementation and use of the tool is clumsy compared to how Lightroom handles this: Both the source and target areas are encircled and can therefore be seen. You can move around the source and target circles and change their sizes.
DPL does hide the source and target areas - what an odd decision!

Have you noticed that using the clone tool on an image with no local adjustments is faster than on one that has them? It makes a huge difference on my older Windows 10 machine with all updates, an i7-6700 @3.4 ghz processor, 24gb of ram and an Nvidia GTX 745. Photolab runs on an SSD drive, while the raw files are on an IDE drive.

When I played around and also hid the local adjustment masks and deselected the local adjustment palette, I had a similar improvement in performance. I wonder if your computer is much higher spec’d than mine so that any performance difference is less noticeable?

Mark

Has not someone from DXO already promised that the handling of source and target selection will be improved in the future? Can’t remember the exact post.

1 Like

Yes, i believe they are working at this moment on this matter.

I have also found some odd areas being used totaly inconstamtly. When I have redone an imige the selecred area changes, I take it the area to be corrected must be slightly diffrent. I had an imige with water, grass and hills behind. At one stage it selected the sky when correcting a poll running through the grass redid and it corrcly went to grass, odd.

@John7 Are you setting the source yourself, or are you relying on DxO?

This was letting it do it’s own thing.

And there’s the problem that the new version of the tool in PL3 solves.

In PL2 the source for repairing was determined “somehow” and we had no control over it. With PL3 we can now determine the source for, not only repairing but, also, cloning. But you need to check the “show masks” option in order to see what is happening. Then you can move the source area to where you want that makes sense for the repair/clone you are doing

See Pascal’s tutorial here

1 Like

Je l’ai essayé. Voici le résultat :

44

One more little feature that I would like to see is the ability to modify the target area’s size and shape.

Or is that just a step too far ?

2 Likes

Too bad.
This very fast operation works on Windows!
Pascal