Musings on Photography (and sometimes, Life :-) )
Barcelona, Spain
[Copyright Andy Richards 2019
All Rights Reserved]
Black and White Image
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
Barcelona, Spain
[Copyright Andy Richards 2019
All Rights Reserved]
[We interrupt our regularly scheduled programming to bring you this “breaking news!” Well, sort of. 🙂 Some just newly-added Photoshop developments have prompted me to post this in-between the continuing story of our most-recent Mediterranean trip. I will get it back on track next week]
IF YOU haven’t looked at the newest editions of processing software, you probably should! Photographic Editing software (we like to call it “post-processing” 🙂 ) has only gotten better over time. I mostly use Photoshop. I see a few of these new “tools” being offered by other software providers. I am sure some of them are very good. And I know from trying them, some are not as good as touted. Because it is what I learned, what I have, and what I have kept up with over the years, I’ll mostly direct my specific comments to Photoshop. I have looked at other software a couple times and if I were starting out, I might well buy into one of the other, really good programs, like OnOne, Topaz or Luminar.
over the past couple years, I have revised my post-processing routine so much, that some of the old photos “begged” re-working
TODAY, THERE is good news and bad news on the “baseball analogy front.” Some of the developments over the last year or two have been “game-changing” for me. But the very newest, splashiest additions? Maybe not so much. Promising. Interesting. Ready for prime time? I am not so sure.
The “engine” for the ACR (Adobe Camera Raw) and LightRoom raw converters just keeps getting better and better
I WILL start with what I think is “the good news.” The “engine” for the ACR (Adobe Camera Raw) and Light Room raw converters just keeps getting better and better. As many of you know, I have written two eBooks on photographing two of the U.S.’s premiere locations for fall foliage photography: Vermont and Michigan (you can find them by clicking on the link for “Photo Destination eBooks” at the bottom of this page). Last fall (2022), we published our 3rd edition of the Vermont Book, which is co-written by my good friend, Carol Smith. In so doing, we added many new images. But in addition to new images, over the past couple years, I have revised my post-processing routine so much (primarily because of improvements and innovation in processing software), that some of the old photos “begged” re-working. And as I did that, I discovered some older, never-processed photos, that – under the new engine for Adobe Camera Raw (ACR), and a couple new Photoshop features – I now felt were worth processing. These were images that I had considered not worth the time, but something told me to keep them in the archives in case new ability came up in the future – and it certainly has. I am not always the smartest or most forward-thinking guy around, and I can think of an old-tech (think film) image or two that I no longer have that I regret not maintaining. But in this case, I feel that my archiving of virtually all my digital files has perhaps been prescient. This month, Adobe has released a couple more digital tools that will undoubtedly have me going back to look at old stuff once again, though as foreshadowed: perhaps not as completely enthusiastically. More in a minute . . . but first:
But how much better and more convenient it is to now do most of that work now right in ACR, before conversion?
MONTHS AGO, Adobe quietly upgraded the ACR (Adobe Camera Raw) “engine,” improving the way masking works in raw image processing and adding and/or improving some of the tools in the engine. For many years, I used ACR basically only to convert raw images, taking them into Photoshop for all other work. I did all my detail work, including masking, on layers in Photoshop after conversion and importing of the raw file. But how much better and more convenient it is to now do most of that work now right in ACR, before conversion?
I am using the ACR masking a lot these days for things I used to wait to do in Photoshop
THE NEW “AI” tech in the raw converter engine crosses all of the Adobe platforms (ACR, Light Room and Photoshop). For example, the new masking algorithms seem to have coincided with (or happened shortly after) Adobe added the “Select Sky” and “Replace Sky” tools to Photoshop, and I find them both (in ACR and Photoshop) to be awfully effective (the “Select Subject tool” in both is not quite as effective, but is still pretty good, and can save a lot of time when making selections in Photoshop – I basically do not use it in ACR). I am using the ACR masking a lot these days for things I used to wait to do in Photoshop. Perhaps my favorite and most useful addition is the sky masking feature in the ACR masking module. I have – in the past year – only found a couple situations where it didn’t do a great job. And in most cases, I have been able to touch up any “near-misses” with the “add” and “brush” features. In the module, there are options to “invert,” “duplicate and invert” and to add or subtract to this mask. I probably use this ACR feature on every image where there is a sky, to make at least two masks (sky, and “not sky” for lack of a better description). I then make adjustments on those masks in ACR, before bringing my image into Photoshop. I find that 85-90% of the time I don’t do any more adjustments (other than sharpening, sizing, profile conversions, etc.) in Photoshop.
IT IS pretty common in my own experience that I want to apply adjustments to a landscape image which I do not want to affect the sky (and vice-versa). The ACR masking tools allow me to easily and intuitively make separate adjustments to individual parts of an image. With an appropriate image (one with a sky), I separate those two elements (sky and foreground) early in the process of raw conversion. I use the mask function to mask the sky and then invert the mask (in my experience, this works infinitely better than using the mask “subject” feature). Then I make adjustments that I want to make to everything but the sky. I love the “shadows” adjustment, the “texture” adjustment, and the “dehaze” adjustment. On first look, “dehaze” and “contrast” adjustments appear very similar. But there are most certainly subtle differences. I generally work those two together, comparing results of each.
TWO OTHER adjustments I am at least considering quite often are “Exposure,” and “Clarity.” It is not unusual, to my eye, that the non-sky portion of my images could use a bit of increased exposure. On the other hand, it is equally likely that I may want to decrease exposure a touch to the sky. I also sometimes will apply a dehaze, and negative texture (among other things) adjustments to skies. There is a pretty cool shortcut for this. I first mask the sky and then, on the drop down, there is choice: “duplicate and invert mask.” In this way, I create two separate masks of the sky and foreground that I want. In an earlier post, I may have “disparaged” clarity a bit. I read a quote somewhere by someone I respect as a digital processing expert, where he made the comment that nearly every image will benefit from a small clarity adjustment. As I work with ACR more and more, I agree. But we are talking a very subtle adjustment in most cases: between maybe 4% and 10,% leaning toward the lower number. Any more than that, and to my way of thinking, clarity introduces what I characterize as a “grunge” look. There are, of course, exceptions (in nighttime, very low light, and severe overcast conditions, I often push clarity harder and feel that it improves things. It is useful to know that both clarity and dehaze are variations on contrast adjustments that each have their own effect. I am going to say these features are the equivalent of a “home run.” 🙂 Now for the “bad news” (or maybe at least “not so good” news).
The real eye-opener is Adobe’s new Generative AI tool, based on its “Firefly” technology
THIS MONTH, Adobe pushed out another significant update: version 24.5. I ask the question whether these are the baseball equivalent of an “error.” That is really pretty harsh. While I hoped the title would grab your interest, I would not really characterize them as baseball “error.” But neither are they a “home run,” in my estimation. Maybe more like a base hit (which doesn’t presuppose a score or a win).
THE MAY updates contain several “improvements.” I think two of them are particularly significant. They are both “AI” based ” features, and work along the lines of Adobe’s other “content-aware” offerings. These additions, by the way, are in Photoshop itself, rather than in the raw converter. This immediately means they are operative only on images that have already undergone conversion from raw to some other image format. One is the automatically included (for CC subscribers) new “Remove” tool. Nested with the Healing Brush and Spot Healing Brush Tools, it is yet another “content-aware” tool that might be more convenient and useful than some of the others.
BACK IN 2009, I traveled to Minnesota to visit my buddy and fellow photographer, Rich, and to meet my good friend, talented photographer and writer, and Photoshop “Guru,” Al Utzig. We spent a fun, long weekend on Minnesota’s vaunted “North Shore.” Perhaps the highlight of that trip was our visit to Split Rock State Park, and the Split Rock Lighthouse. We were there for one of the infrequent “Lighting” events (the lighthouse has been decommissioned) for this spectacular, cliff-borne lighthouse. Because of the event, there was an unsightly (photographically) white “event tent” in the front of the light. Al suggested we go ahead and shoot away, and he would show us an easy way to eliminate the tent in post. That night in our hotel rooms, Al introduced us to the then newest AI technology by Adobe: “Content Aware Fill.” While it didn’t/doesn’t always work flawlessly (more in a minute), all you did was drew a rather loose lasso selection around the offending element (the tent in this case) and used the fill command (keyboard short cut: shift_backspace, and then enter) and poof! the tent was gone. I was awestruck and it became one of my new best friends. For the first time ever, I could calmly shoot an image with people or other things in it, knowing I could “fix” it later.
SINCE THOSE “Content Aware Fill” and companion “Content-Aware Move” tools were introduced, they have improved it over time, doing things like adding a window that opens to allow adjusting the masking and sensitivity of the AI mask used to remove the element. Then later, they added a “content-aware” feature to the crop tool. I have used that liberally, too. But they don’t always work the way you expect (or hope). Sometimes they work flawlessly, but when they don’t, there is still a fair amount of “healing,” “spot healing” and clone work in some cases. The new “Remove” tool applies the content-aware technology to a brush which you can use to “brush out” an offending element. I have barely played with it. But I have done enough to tell you that – like the other content-aware tools, it doesn’t work flawlessly. I am still learning about fine-tuning its use. But I think it has a place in the toolbox. In the Grist Mill Image, above (our trip rendered mostly horrible shooting conditions, unfortunately), we were dodging rain, and I knew my opportunities to get a “pristine” foreground would be limited to non-existent. I shot anyway, knowing I would be able to remove the other people using the “content-aware” technology. Sure enough, in post, I used the content-aware fill tool to remove the three people. Because there was a fair amount of conflicting detail around them, a simple lasso proved problematic. I spent more time than I wanted to, fiddling to get them out in a convincing manner. So I was kind of excited to try the newest “Remove” tool. I was pleasantly surprised. Rather than a selection tool, it is a brush (still making a selection, of course). You can adjust the size of the brush, but there is not much else (yet?) you can do to fine tune the tool itself. One think I found disconcerting was that the default behavior of the brush is that it performs its action after you release the mouse on each stroke. I did discover that there is a check box for that behavior. I leave it unchecked by default now. Many of the objects I want to remove are larger than the brush I am using. I want to be able to finish my brush selection before committing to the action. Of course, that adds some mouse clicks. As I got more comfortable with the tool, I found myself toggling the checkbox on and off, depending on the size and situation of the element I was trying to remove. I also like to keep a “History” window open when doing this kind of work, and it became pretty easy to move back and forth until I was happy with the changes. I will say that I think it did a better job than my original try back in 2011. And I think it is a much easier tool to use for certain purposed. I am happy with it as an addition to the other content-aware tools, as it gives me yet another way to approach retouching. So, I’ll call this one a solid “base hit.”
THE REAL eye-opener is (or should be) Adobe’s new Generative AI feature, based on its “Firefly” technology. To be fair, this one is only available as part of Adobe’s BETA version of Phototshop. “Beta,” of course, by its nature, means it is not really ready for release. And as we will see, it is not. If it can be improved enough to do what it is intended to do, however, it will be pretty mind-blowing. Either way, this one is bound to be somewhat controversial. Understandably. No new technology – especially AI-based – is going to come without some discord. And virtually (see what I did there? 🙂 ) all new digital technologies come with the potential for harm, along with great good. This one is pretty interesting. borrowing from (I assume) Adobe’s database of millions of images and image elements, the “AI” here is not just pixel manipulation. Rather, it is moving more toward an actual artistic attempt to create. It basically takes verbal (written) “suggestions” you make and creates (or composites?) an image or image element based on them. For the image here, I simply loaded one of my existing images into photoshop. Looking at the two side-by-side images, you could conclude they were made just minutes apart. But there is only one base image here, made in 2021. Today (2023), I used the new Generative AI tool. I made a selection (in this case a rectangular one) on the image where I thought a bike and rider might look interesting and typed “bicyclist” into the dialog box. Telling it to “generate,” Photoshop created the cyclist in the image (as well as a couple other options). If you don’t like their options, tell it to generate again. And again. And again. Based on some YouTube video examples I have watched, you can be much more specific and detailed with your dialog suggestions and (theoretically) get more refined results.
Why would you want to do any of these things?
BUT WAIT, there’s more. You can start with a blank document and let Photoshop generate an entire scene. You can also add canvas to an image and let photoshop contextually generate content (just leave the dialog box blank and tell it to generate). And, by selecting elements in an image that you don’t want in the image (think people), and generating (again with no text), I watched it do a pretty clever job of removing those elements. Why would you want to do any of these things? One of my buddies quipped a year or so back on an email exchange that we might as well just sell our camera gear and sit in front of our computers. 🙂 Obviously, that would remove the primary element that attracts us enthusiasts to our “sport.” I will always go out and make photos. I enjoy the trip as much as the result. But I can certainly see some use for this crazy new technology on the creative side of things! At this point, if you are a Photoshop user and have gone into your PS to find this Generative AI tool, you may be mystified. Where is the thing? I said it was nested with some of the other “Content-Aware” tools. Right? Not yet. This is in Beta. So how did I get it? You have to be a CC subscriber and you have to download the beta version. I did that. The nice think is it loads separately from my already tried and true Photoshop version.
DOES THIS stuff work? Kinda-sorta. If the scene is too complex, or the request too specific, then not so much. If it is a simple scene like an element with lots of water surrounding it, the Remove tool works pretty well. But I have played enough to see that it will only work in certain situations. It placed the bike and rider nicely in the Dunedin Trail scene. It also does pretty well with undetailed, “bucholic” scenes like the one below. The base image is a sunset I have photographed many times near my home here in Palm Harbor Florida. The land mass you see in the background is actually an island out in the Gulf. There is no lighthouse there. It seems to me the Generative AI tool did a decent job of creating a lighthouse to my “lighthouse” verbal prompt (after me selecting an area in the image where I wanted the lighthouse to be).
YOU NEED to plan for scale and position, though, because while the tool seems to do a pretty good job of blending the new element with the color and light surrounding it, you cannot move it without penalty. It doesn’t adapt or re-blend automatically (maybe somebody much smarter than me can figure out how to do that. One thing I did on another image was make the Generative AI addition on a separate layer and then use a soft black brush on the mask. That kind of worked on the Gauley River image where I used Generative AI to add “kayakers in red kayaks.”
FROM THERE, things kind of go downhill. From my limited experience, this “AI” software really doesn’t do a very good job with anything detailed. The kayakers look cartoonish, and the red unnatural. When you zoom in and look at virtually any face I have seen it generate, they are just short of (and in some cases absolutely) horrible. Animated subjects (I have tried dogs, tigers and horses so far) are cartoonish and undetailed. Also, the blending from whatever the source is not always good. Arms and hands are often cut off. Sometimes remnants from (I can only assume) the images from which the generative elements were taken from are inserted in the image. Faces are – again – not well done and often cartoonish, or even unreadable. Color matching can become problematic, especially when brighter images are involved. And sometimes, it doesn’t even come close to what you “verbally” requested. When I tried to get it to place Texas Rangers on the cobblestone walkway in front of the Capitol Building in Austin, it was totally confounded. The words “Texas Rangers” got me either a white bench, or white dogs. “Texas Rangers on horseback” yields only horses. I may need some more tutorial work on this feature. I next tried a large waterfall shot. I tried to place a canoe coming over the edge of the falls. It couldn’t differentiate between canoe and kayak. It does colors well (I asked for green and for red and that came in correct each time). But in all but one try (and I made quite a few), the canoes/kayaks were facing the wrong way. Even when I told it “facing viewer,” or “coming at viewer.” Even when you do get a “decent” generated result, the color, scale, and perspective often needs work. This means making layers, trying to use transformation, adjusting brightness, saturation and color balance. At that point, I think I could find my own image (e.g., of a kayaker in a red kayak), and just make my own composite image. So, as I said earlier, for the most part, not ready for prime time.
From my limited experience, this “AI” software really doesn’t do a very good job with anything detailed
THERE IS one area where the Generate AI feature seems to fare a bit better: when you give it context to work from. In the image below, I pretty much hacked up the original composition. Not sure what I was thinking. I thought it might have been salvageable if I hadn’t cut off the kayaks in the middle, leaving the rest totally out of the frame. I cropped the image some, and then, continuing to use the crop tool, I extended the left edge out enough to make room for the “rest” of the kayaks. It took me a couple tries to get the extended portion wide enough for Photoshop to figure out I wanted the entire boat in the image. I then selected the blank canvas and told Photoshop to “generate” without giving it any verbal prompts. I am led to believe that this causes the software to interpret the suggestion to use the surrounding image for context. I was reasonably impressed with the result.
IN THE Chicago image, I tried using Generative AI to remove people from an image (not my idea – got it from a YouTube review). It worked better, at least for my one example. But it still didn’t get it right on the first try. I had to use Content Aware Fill to remove remnants of the flag above the policemen, and what the AI “thought” should go in there, which was more flag. So conclusion so far: not a good as I had hoped or expected. But you would expect this to get better over time. Until then, a new “tool” for the bag, to be used sparingly. I will still be relying on those tools that consistently work for me. For now, Generative AI is kind of a weak, ground ball. It will take a combination of circumstances and luck for the runner to even get on base. But I do say “for now.” This AI stuff – scary and wonderful at the same time, will keep on coming and keep on getting . . . well, more “AI.” 🙂
For now, Generative AI is kind of a weak, ground ball
IF YOU are a CC subscriber, you should be able to find the beta Photoshop version in your Creative Cloud software. There is a “beta” slider you can use to download it. It doesn’t load into your existing setup. It loads separately. Find it. Play with it. decide for yourself. But you do, in my view, owe it to yourself to take a look at some of these new features – if you haven’t already done so.
(Parc Naturel Region de la Sainte-Baume)
[Copyright Andy Richards 2022
All Rights Reserved]
(Parc Naturel Region de la Sainte-Baume)
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
Parc Naturel Region Sainte-Baume
[Copyright Andy Richards 2022
All Rights Reserved]
Aix-en-Provence, France
[Copyright Andy Richards 2022
All Rights Reserved]
France
[Copyright Andy Richards 2022
All Rights Reserved]
France
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
FTER WALKING around some, we sat in a small cafe and had coffee and wine, before meeting with Franck at the exit from the feudal village to head back to Toulon and reboard our ship. A very nice day. It was just a taste, but I can see that the Provence region is a place that would merit a few days’ stay and exploration.
[Copyright Andy Richards 2022
All Rights Reserved]
Monaco
[Copyright Andy Richards 2022
All Rights Reserved]
Monaco
[Copyright Andy Richards 2022
All Rights Reserved]
Monaco
[Copyright Andy Richards 2022
All Rights Reserved]
Monaco
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
[Copyright Andy Richards 2022
All Rights Reserved]
LightCentric Photography Blog 