Photoshop 2020, Luminar 4, AI editing, and Sky Replacement

[Thlayle]

I think this was prompted by Manfred's comment in another post about the latest Photoshop CC update (to the effect that his money used to purchase Luminar 4 (Skylum Photography) seemed wasted since sky replacement is now a feature of Photoshop). Also, in my last forum post, I brought up the question of whether or not, PS will keep up with the newest AI developments, such as Luminar's AI sky replacement which as on the market about a year ahead of PS sky replacement (the response was underwhelmingly silent -- no comments from anyone).

Now, having tried PS sky replacement and comparing to Luminar 4, I will say that it is not wasted money to purchase Luminar -- though right now make sure it is Luminar AI, not the Luminar 4 product.

Luminar 4 and PS sky replacements work about the same. But it is Luminar 4's other AI tools that make it far ahead of PS for sky replacements. With its other AI editing features, it makes for a much, much easier and quicker makeover of an image.

The attached image is a quick edit I did, using Luminar 4. It could be better of course, but to do it better...well PS isn't necessarily the answer.

I am excited to see what the new Luminar AI will do. It is an expansion of their AI editing approach.

Like it or not, AI editing is here to stay.

For my part, I am going with it. Still not seeing it as a PS replacement for all things. Still, Skylum Software is probably going to give Adobe a run for their money.

Regards,

Randy

[purplehaze]

Here's a response* from one of my favorite photo vloggers, Thomas Heaton:



Sorry I don't have time to locate the exact start of his critique, but I think he is worth watching anyway. If I can garner him some more fans, so much the better.

*Sorry, it's not a response to the question posed, but to the role of AI processing generally, which as someone who has been playing with Luminar 4 and compiling a bank of my own sky photos, I have been wrestling with.

If you think it is better placed in its own thread, maybe one of the moderators can do that.

(Backing out of the room in a bowed, apologetic stance.)

[Manfred M]

Randy - the two products are different and I personally find that the Photoshop implementation is much more flexible (and powerful) than the Luminar one. With Luminar, the system dumps out a finished product and fine tuning the results is challenging. In Photoshop, everything is output as a set of adjustment layers, which allows the user to fine tune the results that the algorithm has created.

If one wants an quick and easy sky replacement, then Luminar is fine. If one wants more control to fix areas where the algorithm give results that are less than ideal, then Photoshop is much stronger. In my view, would I spend the extra money with Luminar now that Photoshop has comparable technology? In my case the answer is clearly no.

[Thlayle]

Thanks. I haven't watched the whole thing, but when he got to his complaints about Luminar AI I found him quite tedious.

Everything he did have to say about the flawed results and the absurd complaint that the software does not give results that are "real" -- those all are comments that can be said of PS and pretty much any post-processing of digital images. (Needless to say, you have to made me a fan of him yet. Maybe I will try some of his other posts.)

I think he cherry-picked the images with flawed PP, but even if he didn't it should be no surprise that sometimes post-processing mistakes happen. Like any software, this can go wrong, so I think he really misses the point.

As for the lost realism he refers to: that is totally laughable. In one of my first posts at the CIC forum, I asked if members had any reservations about losing sight of "realistic" results and the whole question of honesty. CIC members responding told me my concerns were misplaced and at times their comments bordered on ridiculing my concerns. So I am having a lot of trouble with this complaint.

He also refers to Luminar AI -- not available yet. So unless he's using a beta version, something's wrong with his whole post. If it is a beta: maybe that has something to do with some of the flawed results he showed.

AI works through the same tasks that are already out there to be done, but does them much faster. This is another idea I have heard cherished again and again in this forum: the need for speed; how if you are a serious photographer or a pro, you can't waste any time. So there's that. AI improvements stand to improve efficiency of effort, another point missed in the vlog.

I am not a big fan of sky replacements an similar alterations, but much of my shooting takes place when I am following an itinerary that is NOT centered around photography, such as when traveling or on the spur of the moment when a unique opportunity presents itself. AI sky replacement and other pp changes can rescue otherwise bland landscape shots. Having compared the two products, I say Skylum's software fills a specific purpose, and it does it better than PS (IMHO).

And I love Unmesh Dinda's vlogs. He covers a lot of depth in short bursts, and he really explains things really well.

My point in all of this that the Skylum software is doing at least one task that PS does, but I would say it is even better at it. It's worth a try if the price tag (about 64 US $, I believe) doesn't seem too much.

On the broader topic of "is this good" for photography, I have the say yes I think so, at least as good as other digital pp tools.

[Thlayle]

Not sure what you mean about being sorry and backing out of the room in an apologetic stance. I wasn't posting a question as much as suggesting an idea to consider. So no worries, I see no call for apologies. But I did find the vlog off on this subject. His banner on the video "Never Wait..." seems misleading, given his take on the software.

[Thlayle]

Randy - the two products are different and I personally find that the Photoshop implementation is much more flexible (and powerful) than the Luminar one. With Luminar, the system dumps out a finished product and fine tuning the results is challenging. In Photoshop, everything is output as a set of adjustment layers, which allows the user to fine tune the results that the algorithm has created.

If one wants an quick and easy sky replacement, then Luminar is fine. If one wants more control to fix areas where the algorithm give results that are less than ideal, then Photoshop is much stronger. In my view, would I spend the extra money with Luminar now that Photoshop has comparable technology? In my case the answer is clearly no.

Thanks, Manfred. I am not thinking of this as successor to PS, not in the least, although I think for many users it may have way more curb appeal.

I can't agree though that PS does sky replacement as well as Luminar 4. I think for the basic task of replacement, it does about the same. For the final result, getting the sky replaced and the overall look right, I found Luminar much MORE flexible, not less. The trick seems to be getting it right before hitting the "apply" button (dumping as you would call it). After that, I agree that going back and fixing flaws is a lot of trouble and I will defer to you judgment there that PS has more power in that respect.

So for my part, I am not sorry I bought this product. For one thing, there's many a situation where in fact I do just want a quick result. I am not a professional photographer, just an amateur with a hobby. Still, I value my time, so like the pros who have commented in this forum at times: the speed is worth a lot. In terms of results, no I don't think PS has the flexibility to do this simple task a well or as quick. At least that's where I'm at on it right now.

[Manfred M]

Here's a response* from one of my favorite photo vloggers, Thomas Heaton:


Sorry I don't have time to locate the exact start of his critique, but I think he is worth watching anyway. If I can garner him some more fans, so much the better.

*Sorry, it's not a response to the question posed, but to the role of AI processing generally, which as someone who has been playing with Luminar 4 and compiling a bank of my own sky photos, I have been wrestling with.

If you think it is better placed in its own thread, maybe one of the moderators can do that.

(Backing out of the room in a bowed, apologetic stance.)

Janis - I had a quick look at the video you posted and I understand his point. It is one that has been around for as long as photography; should we rework our images and if so, what limits (if any) should there be? This goes the whole gamut from not touching the image at all and living with what the camera has recorded to getting in there with modern tools. When people ask me if I use Photoshop in my work I will often get the knowing smile that I haven't got a "real" image and that I must have somehow cheated to create it.

These ideas are not at all new and almost as old as photography itself. I'm currently reading Beaumont Newhall's classic work "The History of Photography" and in it he points out that photographers, even in the 1800s were "enhancing" what their camera captured. In the film days, photographers used specific films and papers that had certain characteristics that they liked. Different film emulsions, both B&W and colour, created different looks. Different camera systems and lenses created different looks. At the very high end, Zeiss was known for producing lenses that had a clean, crisp, "surgical" look whereas their main competitor Leica produced lens that were incredibly sharp with an almost creamy bokeh. In the digital age, people will talk about the Canon or Nikon or Sony, etc. "look", which is really the result of engineering decisions made by the camera manufacturers.

AI techniques (or perhaps more correctly, using machine learning) are just another a tool that can be used or misused, depending on one's perspective. I have used sky replacement from time to time, but only when I have no other option; i.e. the sky is so bad that I can not pull a good image out of the capture. In much of my more recent work I try to capture images where the sky is purposefully excluded. I think we are getting into the same issues that we saw with HDRI, long exposure (especially when shooting water and clouds), partial colourization and many techniques that have been abused and have developed a bad reputation among "serious photographers". Unfortunately, a poorly made image cannot be improved just by throwing some of these techniques at them; the image is still sub-par in the end. On the other hand, well done image that use these tools and techniques well will continue to be stunning images. If you look at most of my recent work, virtually the only tool I tend to use is classic dodging and burning; like the old masters like Ansel Adams did.

The bottom line is that cameras do not "see" the same way that humans do, so interpretation of the image to tell our story is certainly a legitimate approach. There is no right or wrong answer here, but personal taste is certainly a key factor. My personal taste / style tends to be vibrant and punchy images. Not everyone likes my work, but enough people see it for what it is.

[Manfred M]

I am not a professional photographer, just an amateur with a hobby. Still, I value my time, so like the pros who have commented in this forum at times: the speed is worth a lot.

Carefully said, "pros" covers a lot of ground. I know quite a number of photographers in the business and the amount of time they spend on an image ranges from seconds to minutes to hours to days.

Wedding, portrait and event photographers - for the most part, spend seconds on an image in post. One international award winning wedding photographer I know figured he spends between 20 and 40 seconds on an image. This is probably about the same as what an event photographer I know does.

Portrait photographers I've spoken to are much the same as the wedding photographer, although one who does large prints suggested he will spend 5 -10 minutes making a print.

When you get into commercial photography retouch time on an image is measured in minutes to hours, depending on the client and final use. In a large, high end advertising campaign, hours is definitely the typical time.

If you look at fine art photographers - hours to days is not atypical. One of the high end print makers I've gotten to know quite will used to do Karsh's colour portraits and it would often take a week of test prints to get the look that was required and it would take a full day to do the final print that would be delivered to the customer.

So while time is money. giving the "client" the quality that is required depends a lot on the genre and the clients needs (and budget).

[Thlayle]

Manfred, wow I hadn't thought of that range of situations and nuances. I just remember quite often seeing the comments about the valued time and choice of methods.

I see that Luminar does use layers and editable masks, but it seems to apply them very differently than PS. And there does not seem to be a comparable file format (like PSD for example) that saves the work without having to come up with new masks, layers, and so forth. That does seem to be a real shortcoming of this product.

[Manfred M]

Manfred, wow I hadn't thought of that range of situations and nuances. I just remember quite often seeing the comments about the valued time and choice of methods.

I see that Luminar does use layers and editable masks, but it seems to apply them very differently than PS. And there does not seem to be a comparable file format (like PSD for example) that saves the work without having to come up with new masks, layers, and so forth. That does seem to be a real shortcoming of this product.

The different pieces of software will implement routines differently.

Software like Capture One and DxO Photo Lab uses layers, as does Photshop, Gimp, Affinity Photo, etc. Other pieces of software like Lightroom and other parametric editors do not, so one size does not fit all scenarios. Luminar's newest features require a database / catalog; older versions did not.

[purplehaze]

Janis - I had a quick look at the video you posted and I understand his point. It is one that has been around for as long as photography; should we rework our images and if so, what limits (if any) should there be? This goes the whole gamut from not touching the image at all and living with what the camera has recorded to getting in there with modern tools. When people ask me if I use Photoshop in my work I will often get the knowing smile that I haven't got a "real" image and that I must have somehow cheated to create it.

These ideas are not at all new and almost as old as photography itself. I'm currently reading Beaumont Newhall's classic work "The History of Photography" and in it he points out that photographers, even in the 1800s were "enhancing" what their camera captured.

I don’t read Thomas as talking about enhancement here, but rather, dramatically changing reality, changing atmospheric conditions, puttting mist where there was none, turning midday into dark, that kind of thing. Not that there isn’t a place for that, but is it landscape, or something else? His second point I imagine you would relate to as a judge and has to do with disclosure. Is it honest to pass off dramatically altered landscapes as reflecting reality? His third point, about AI biasing choices away from novelty I think is scarily true of all AI. I wonder about the impact on our appreciation of nature, and aesthetics.

Yes, there is much that is seductive about the tool. But how much will people who only wield it really understand about what they are doing, I wonder. Food for thought.

[Manfred M]

Janis - that is certainly one point of view that many people hold. On the other hand, here is an image I've posted before, by Ansel Adams, his well known Moonrise, Hernandez, New Mexico.





And this is what he started with.




I would suggest one of the most celebrated landscape photographers ever did exactly what Thomas disagrees with. I would also suggest that if we had access to the original of most of his works, we would find these major transformations as well.

There is no right or wrong answer here; just opinion and personal taste. Just to be completely frank, I personally like a lot of Adams's work, but not this particular piece. I understand he made in excess of 1300 prints of this negative over the years.

[purplehaze]

I didn’t know that, Manfred. Thanks!

[Elle Harper]

Beautiful, nothing else can describe, awesome.
https://fixthephoto.com/photodirector-vs-photoshop.html

[xpatUSA]

Janis - that is certainly one point of view that many people hold. On the other hand, here is an image I've posted before, by Ansel Adams, his well known Moonrise, Hernandez, New Mexico.





And this is what he started with.



<>

There is no right or wrong answer here; just opinion and personal taste. Just to be completely frank, I personally like a lot of Adams's work, but not this particular piece. I understand he made in excess of 1300 prints of this negative over the years.

Thanks for the 'before' and 'after', Manfred. I recall reading that He saw the moment, stopped the vehicle immediately and shot with the emphasis on speed of capture rather than perfect composition/exposure.

[Manfred M]

Thanks for the 'before' and 'after', Manfred. I recall reading that He saw the moment, stopped the vehicle immediately and shot with the emphasis on speed of capture rather than perfect composition/exposure.

I remember reading that he managed to get just one shot in before the sun went down.

I know,one photographer who shoots landscapes with a view camera. For him “burst mode” is two shots per hour.

[loosecanon]

Surely the problem with automated sky replacement is when the new sky is not the work of the photographer and thus the image is not aall the work of the photographer, a requirement of most competitions etc.

That said the PS sky replacement tool is a quick and simple way to obtain a mask which can be used by the photographer to add their own sky into an image. All those trees nicely masked.

[xpatUSA]

Surely the problem with automated sky replacement is when the new sky is not the work of the photographer and thus the image is not all the work of the photographer, a requirement of most competitions etc.

Good point!!

While learning layers/masking I once posted a composite which had a "stolen" moon in it. Fortunately, I had replaced that layer with my own better super-moon shot before I entered the final composite into a mini-competition here ...

[Manfred M]

Surely the problem with automated sky replacement is when the new sky is not the work of the photographer and thus the image is not aall the work of the photographer, a requirement of most competitions etc.

That said the PS sky replacement tool is a quick and simple way to obtain a mask which can be used by the photographer to add their own sky into an image. All those trees nicely masked.

Agreed, but the automated sky replacement tools allow the photographer to select the sky that they use, not just the library of skies that are built in to the software. Replacing a sky with one I have taken is generally permitted in any competition where editing is allowed. Competitions where editing is restricted would likely not permit sky replacement.

[PhilT]

Surely the problem with automated sky replacement is when the new sky is not the work of the photographer and thus the image is not aall the work of the photographer, a requirement of most competitions etc.

That said the PS sky replacement tool is a quick and simple way to obtain a mask which can be used by the photographer to add their own sky into an image. All those trees nicely masked.

You are just writing about a point I wanted to ask: I don't use too much PS, being just an amateur and doing most or all the PP with LR. But I recently tried the PS sky replacement on several pictures with dull skies and I liked the results and the simple "one click and it's done" But I felt that the choice of skies is limited and wondered if there is a way to use my own "skies" instead with it, or adding them to the list of the PS choice.
If it's possible could you explain to a very basic PS user how to do it?
Thanks a lot if you can
Philippe