Processing in AI (split-off thread)

[michali]

Advertisement (gone after registration)

On 9/12/2025 at 9:28 PM, chris_tribble said:

 It tried it.  Does it maintain the dng or produce a tiff?

 

On 9/12/2025 at 9:44 PM, jaapv said:

I use it in Photoshop as plugin at the very beginning of my workflow when opening from ACR so I would say .psd;  this question does not really come up. Even more so in LR, as the DNG conversion takes place during export, LR being non-destructive.  AFAIK Topaz standalone works with TIFF (but I may be mistaken), unlike DXO which provides the best noise reduction and exports as a new separate DNG.

I use the standalone Topaz DeNoise as the last step in my workflow i.e. after converting DNG to TIFF in PS & then import the TIFF into Topaz. I use the noise reduction/sharpening very lightly & sparingly & only for images like the ones below. 

FWIW- I see that the Topaz DeNoise standalone is no longer available, it's now included in their Photo AI which I somehow bought & was extremely disappointed, esp. at the way it over-manipulated images, even at the lowest settings - a complete waste of USD200-  😠 

(please click on images for better res.)

SL2 & VE90-280mm @ 144mm  ISO12 500

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

SL2 & Sigma 150-600mm L @ 600mm, image cropped to 80%   ISO12 500 

 

[jaapv]

I find it useful as a very first step to clean the raw conversion up. It does indeed over-exaggerate any edits if used at the end. 

[michali]

13 minutes ago, jaapv said:

I find it useful as a very first step to clean the raw conversion up. It does indeed over-exaggerate any edits if used at the end. 

Thanks Jaap interesting to note.

I must admit I'm not much of a post processing expert. I used LR in the past when I had more time on my hands, now use PS Elements, mainly time constraint related.

For images I print I use LR.

 

Edited September 14 by michali

[jaapv]

I hardly ever use LR, it does not mesh properly with my mind. I’ve been on Photoshop for two decades now. 

[hansvons]

10 hours ago, pgh said:

I never use de noising for any of my cameras at all - including my sometimes frustratingly noisy in some ways SL2. Oh well.

Neither do I, except for a minimal denoise in the colour channel. This leaves the noise as it is and makes it look a bit like grain. 
—-

Totally agree on your philosophical insights, universal truths etc. AI does have its place but not in photography—a least if it considers itself as authentic. 
 

BTW, I consider texture an essential part of photography. It provides the eyes something to hold on and adds natural sharpness to the pictures.

[pgh]

11 hours ago, jaapv said:

TBH I have little interest in in the philosophical aspects of the process and  tend to judge the result

I can’t get to this place as the process inevitably informs the meaning of the result. For me, the meaning of a work is a necessary part of being able to decide if it is successful or not. 

[jaapv]

Advertisement (gone after registration)

3 hours ago, pgh said:

I can’t get to this place as the process inevitably informs the meaning of the result. For me, the meaning of a work is a necessary part of being able to decide if it is successful or not. 

Given that the whole process of creating a photograph is an artificial mechanical/electronic ( chemical ) one, the nature of the tools used will always be of no philosophical interest to me. Everything we do is manipulation.  The nature of the road does not determine the destination. 

[jaapv]

As for AI tools adding data, quite a few postprocessing tools are marked AI when they are fully so, but I am convinced that significant parts of the newer general  programs incorporate AI in their regular workflow, like automatic selecting and masking, Content Aware technology, colorizing and more without advertising it. The enormous advances in those fields in the last couple of years certainly suggest it.  This will become prevalent, I am sure, and will make life hard for those who wish to avoid such technology.

It is unavoidable, without AI it will be next to impossible to handle the exponentially growing need for processing and data power exacerbated by the Megapixel race. For us simple users it will be accept and adapt or be left behind.

 

[pgh]

3 hours ago, jaapv said:

Given that the whole process of creating a photograph is an artificial mechanical/electronic ( chemical ) one, the nature of the tools used will always be of no philosophical interest to me. Everything we do is manipulation.  The nature of the road does not determine the destination. 

It determines the state that you arrive at the destination in, and the journey absolutely impacts the nature of the destination. It’s probably why you go on a safari instead of using mid journey to make pictures of animals. Or why people hike Everest instead of merely hitching a ride up.

You can pick things apart endlessly, and at some point nothing means anything because no truly perfect logic exists. For me, the effort and intent matter - beyond photography that’s the world I want to live in. Not one that is void of inherent meaning. That’s a dark cynical road and as I said one that is quite a cozy help to would be fascists. 

Anyways. We will disagree here and I’m not going to engage further on it.

[jaapv]

Happy to disagree 😀  The world that we live in is not the world that we show in our photographs. It is the bit in between that we disagree on.

[LocalHero1953]

I use AI Denoise a lot, previously with Topaz, now with Lightroom; occasionally I use Topaz for sharpening. The photos that get this treatment are of stage performances, often under poor lighting, so at ISO 2,500-25,000. The processing produces remarkably clean images that allow the eye to focus on the performance. I am not primarily trying to create great works of art, but rather tell the story of the performance by reflecting dialogue and showing relationships - so composition and facial expressions are important. I entirely get @Stuart Richardson's comments; although I could not easily see the unnatural edits he referred to (I shall look again when I have more time), I shall see if I can replicate his findings in my images.

My images appear to satisfy my audience (primarily the performers themselves) partly because they are colourful and sharp (a combination of Leica colours* and mainly non-AI editing), but also because of the detail visible in expressions often at a distance (I may balance a person in the foreground against someone at the back of the stage). Given that Topaz is often over-enthusiastic with its 'Recover Faces' and 'Super Focus' modes I use them with caution. In the end I just use my own judgement about whether the image I have produced is realistic or not - I'm not trying to fake something that wasn't there. Sometimes I might teeter on the razor edge between acceptably real and distractingly artificial. But then I see photos taken by those inexperienced in editing which have been 'improved' by gross over-use of non-AI sharpening tools. Frankly I would rather have my AI-assisted denoised and sharpened images than those.

 

*Edit. Colours - this is a different kettle of fish. Stage lighting is always coloured, traditionally by gels over broad spectrum lights, but increasingly by digitally controlled LEDs which use narrow bands of frequencies. The human eye and brain are remarkably good at thinking skin tones are normal pinky-orange skin tones (for caucasians), even if they are actually magenta, green or a pungent yellow. The trouble is this particular brain function switches off when it sees a photo - and sees magenta, green or yellow skin. I have carefully looked at a live scene to see its 'real' colour, and I know my camera really is recording the true colour ("yes, if I look carefully, I can see her face under that lighting really is a mustard yellow") - but the eye sees it later as the 'wrong' colour. So I adjust the colour in the image - I tone it down, adjust the saturation, adjust the white balance, occasionally changing the colour mix - all to make a fake picture that looks real. (And I'm sure AI tools will come along, if they haven't already, to do this job for me).

Edited September 14 by LocalHero1953

[jaapv]

Recover faces often leads to cartoons. 

[jaapv]

6 minutes ago, LocalHero1953 said:

Sometimes I might teeter on the razor edge between acceptably real and distractingly artificial.

In that case a spot of Gaussian Blur and some added noise will help. 

[chris_tribble]

Sorry for the many typos in the previous post! Errata:

modern = most

 hope = how

🙃

[jaapv]

12 hours ago, hansvons said:

Neither do I, except for a minimal denoise in the colour channel. This leaves the noise as it is and makes it look a bit like grain. 
—-

Totally agree on your philosophical insights, universal truths etc. AI does have its place but not in photography—a least if it considers itself as authentic. 
 

BTW, I consider texture an essential part of photography. It provides the eyes something to hold on and adds natural sharpness to the pictures.

I can understand the objection to AI adding data to one’s image even if I do not share it in this context. However, specifically noise reduction is a different case. AI is used to collect a large amount of data to be able to recognize noise, making it able to remove it precisely. So in this case nothing foreign is introduced. 

[hansvons]

6 hours ago, jaapv said:

AI is used to collect a large amount of data to be able to recognize noise, making it able to remove it precisely. So in this case nothing foreign is introduced. 

I use C1 and have Topaz Photo and Gigapixel (mostly to upscale old photos for campaigns; for unknown reasons they screwed up their products along the way), and the inevitable Photoshop. Yes, it’s astonishing what AI can do to improve images. However, what one sees as an improvement, might for someone else a degradation or even a unwanted interference. I like texture, others want their images to be as clean as possible.

The beauty of AI and conversely the curse is the fact that it is holistically machine-learned and as such every thing it "sees" interprets from exactly this perspective. Regardless of what you do, masks, noise, you name it, the AI will re-compute parts of the image. Whether that‘s OK or not, whether a classic algorithm is less intrusive or not, is all up for debate and ultimately one‘s decision.

Philosophically I‘m with @Stuart Richardson and @pgh when it comes to photography. In terms of fine arts do whatever you want to do. No boundaries and that includes AI.

That said, I use AI extensively (Claude) for my work. It‘s a brilliant editor, assistant, co-worker, researcher, you name it and can multiply one‘s output. But it requires a deep personal (!) understanding of the topic you are working on and what AI can do and can’t do. That‘s called AI-literacy. If you believe that AI makes you smarter or can substitute knowledge you are intellectually doomed.

And herein lies the danger, which @chris_tribble rightfully called out when he criticized students at uni using AI. However, if they are fully aware of what AI can do and what not, it will help them to speed up things without giving up themselves to the algorithm. As many can‘t see that because they are already immersed in algorithms through endless TikTok scrolling etc. and believe that with AI they can outsmart their Alma Mater, which they somewhat do, they get a completely wrong sense of what education and scientific work is about.

The good news is that the original thought cannot be produced by AI—yet. And that’s what art and science is all about.

[jaapv]

Yes, but traditional editing recomputes the image as well. I think that we must differentiate between functions. When AI is used to compile large amaounts of data to refine a tool (select an object, recognize noise, etc) it is not recreating the image. And doing nothing else than the same pre-AI tools we had, only better.
If it is using AI to create bits of the image that never were there we have another situation but we accepted non AI tools that did the same before, like content aware cropping, cloning, dust removal, etc. We also accepted techniques that changed content away from reality, like changing perspective, both vertically and horizontally. We even accepted techniques like removing an object and filling in the background. My point is that we suddenly draw a line because of a bee-sting acronym AI, without considering what it is doing and how that relates to accepted methods. 

[LocalHero1953]

A year or two back I toured the WW Winter photography shop in Derby, a business that started in the C19; on its top floor is a gallery from the same period with a row of desks beneath a row of skylights, where the photo retouchers worked with brush and ink on glass plates. The retouchers didn't even know what the people in the photograph should look like, unlike those who use AI tools now. It was a big business back then.

I realise I'm just stirring the pot with this, but maybe it helps each of us work out exactly where our own personal dividing line is between acceptable and unacceptable. Perhaps it's the difference between a photo intended as a representation of what was there (landscape, wildlife, reportage), and a photo intended to say something (a portrait for recollection, story telling, advocacy, advertising).

Edited September 15 by LocalHero1953

[pgh]

3 hours ago, jaapv said:

Yes, but traditional editing recomputes the image as well. I think that we must differentiate between functions. When AI is used to compile large amaounts of data to refine a tool (select an object, recognize noise, etc) it is not recreating the image. And doing nothing else than the same pre-AI tools we had, only better.
If it is using AI to create bits of the image that never were there we have another situation but we accepted non AI tools that did the same before, like content aware cropping, cloning, dust removal, etc. We also accepted techniques that changed content away from reality, like changing perspective, both vertically and horizontally. We even accepted techniques like removing an object and filling in the background. My point is that we suddenly draw a line because of a bee-sting acronym AI, without considering what it is doing and how that relates to accepted methods. 

Just a point of clarification. There are certainly segments of photographers that did not accept some or all of these things. The “we” is assumptive and imprecise. 
I also reject the idea that people don’t consider what it is doing or how it relates to methods. In fact it is exactly with these considerations in mind that some choose to reject most or parts of the uses of the tech. 

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[jaapv]

I am certainly not being dismissive of other opinions, “we” was not meant as a negative about other opinions, just a generalization. Sorry if it reads that way.
There are also photographers who dismiss all postprocessing. Or shoot JPG only, photography is a broad church. 
However, I try to define the point at which we draw the line. AI as a creative tool is not my cup of tea, but I think that the acronym AI must be used clearly. Does it refine  tools or does it use big data to creatively alter, modify or even generate the image? Two wholly different things.  My question here is: Noise removal is an accepted image modification. If we accept the blurring method, why dismiss the use of a more precise tool?