I Let ChatGPT Edit My Photos in Lightroom !! Watch

[Krish Chandran]

Oh...lest this comment be dismissed as NOT being a discussion on the specific use of AI in Lightroom and maybe photoshop, let me add this :-

(1) the output shown in the youtube video is kinda least I'd expect from "generative AI" as it stands today. In fact rather underwhelming I'd say.

Not sure what you mean by this or what you expect. Generative Ai alters the image - which is one of the big complaints against it.  What I do know is that by adjusting sliders in Lightroom no generative AI is being deployed. AI is moving those sliders instead of a human. The integrity of the RAW file remains untouched.

(2) About the bullet points -

• No pixels are being artificially generated or swapped out dynamically - Why would it do that if its preserving the RAW file ? And if it did that how would you find out ?

RAW editing is done in Lightroom. After edits the file is still an ARW/CR3/NEF file. Still RAW.  Pixel level edits are done in Photoshop where the files become PSD on import. This person has done another video where he has tried the same thing in Photoshop where Pixel levels edits are made on PSD files. This video only covers edits on RAW files in Lightroom.

[/color][/b]
• The AI actively operates your actual Lightroom software, mechanically adjusting the sliders and creating precise masks - so did our good old hands ! The sample chosen is a standard underexposed image that gives a lot of editing latitude. What if it were the same image, correctly exposed, but with an added artefact (say a cat or a mouse) that needed to be highlighted and treated locally unlike the edited files the AI has based its GIGO learning upon ?

Your good old hands were and can do it. When your images are many - in the 50s and 100s your good old hands and eyes and brain gets tired. So you take breaks and the whole process takes time. This process cuts time by applying changes quickly and accurately. Reminds me of the argument : "we have legs, we can get where we want - what's the use of cycles or cars?
The edits are all local - shown in the video. Whether it is a cat or a mouse or a bird or a mountain or waterbody once you show AI how you want it edited it will do so. The trick is in making the correct prompt. Everything hinges on that. Since you seem to be so knowledgeable I was rather hoping you could shed some light on that

• It preserves the full quality of your original RAW files exactly like a manual workflow would - so whats different/new ?

It is not the difference but the sameness that is of value. At the end of an edit your RAW files are exactly as if you edited them in your style, unlike current day batch edits which are quite hit or miss. 

• Every single setting remains completely adjustable, allowing you to fine-tune the final image 90% of the way through. You can still add to the edits to finalize the end result. - Didnt batch processing take us 75-80% there, albeit with the added benefit of user deciding which template to apply to each batch ? What is the level of incremental gain here compared to the cost (allowing computer control) ?

Have you ever tried batch edits with masks? Invariably after image no 3 or 4 masks shift away from where it was applied in the first image. So the answer to that was to batch edit all non masking edits and then go back and apply masks one at a time. Laborious and time taking. In this video he shows how each mask is individually applied to each image. Time and effort saved.

To me the video is more of a 'clickbait' experiment which demonstrates what most of AI supporters (learned or otherwise) and shunners (also learned or otherwise) already know.

Then why bother with all that you wrote above?  We have all seen clickbait . . Do we respond to clickbait ?
You also said that you were "underwhelmed" by this video yet you made a half page of questions in addressing them in a manner as if I have the responsibility to defend what is shown in the video - that must have taken some effort. For what?

 

In all fairness, We All also know the standard arguments regarding the "AI is panacea to all our problems" and the perceived urgency to modernize our traditional post-processing skills, how AI can help fly you to the moon and back and so on...  let's leave the prophetic predictions aside for this instance. If you could just plainly tell us what you want us to write, I'm sure some of us will simply go ahead and do it ! At the moment the "approved band" for responses is so narrow, most of us barely feel confident of making it.

#I do not know of any one who said that AI is the Panacea of anything - except you - just now
#There is no "urgency" to modernize PP skills. It is moving ahead all on its own. Grumbling/ranting  will not slow it down
#What I wanted is to talk about what this means for people like me and others who spend a fair bit of time in post processing in Lightroom and if and how we could implement it in part or whole or if any one else knew of such a thing. Instead of constructive conversation this has become a confrontational exercise in Ai bashing or Ai messenger bashing.  I should not be surprised, I guess

[SwanOfUtrecht]

In all fairness, We All also know the standard arguments regarding the "AI is panacea to all our problems" and the perceived urgency to modernize our traditional post-processing skills, how AI can help fly you to the moon and back and so on...  let's leave the prophetic predictions aside for this instance. If you could just plainly tell us what you want us to write, I'm sure some of us will simply go ahead and do it ! At the moment the "approved band" for responses is so narrow, most of us barely feel confident of making it.

Oh...lest this comment be dismissed as NOT being a discussion on the specific use of AI in Lightroom and maybe photoshop, let me add this :-

(1) the output shown in the youtube video is kinda least I'd expect from "generative AI" as it stands today. In fact rather underwhelming I'd say.

(2) About the bullet points -

• No pixels are being artificially generated or swapped out dynamically - Why would it do that if its preserving the RAW file ? And if it did that how would you find out ?

• The AI actively operates your actual Lightroom software, mechanically adjusting the sliders and creating precise masks - so did our good old hands ! The sample chosen is a standard underexposed image that gives a lot of editing latitude. What if it were the same image, correctly exposed, but with an added artefact (say a cat or a mouse) that needed to be highlighted and treated locally unlike the edited files the AI has based its GIGO learning upon ?

• It preserves the full quality of your original RAW files exactly like a manual workflow would - so whats different/new ?

• Every single setting remains completely adjustable, allowing you to fine-tune the final image 90% of the way through. You can still add to the edits to finalize the end result. - Didnt batch processing take us 75-80% there, albeit with the added benefit of user deciding which template to apply to each batch ? What is the level of incremental gain here compared to the cost (allowing computer control) ?

To me the video is more of a 'clickbait' experiment which demonstrates what most of AI supporters (learned or otherwise) and shunners (also learned or otherwise) already know.

[Krish Chandran]

This sort of lament is exactly what I was hoping to avoid. It seems one has only to say the word "AI" and all the Thread Hijack ghosts come out of the closet.  Let's stick to any technical inputs that you may have (if any) about how the man in the video made his bulk edits  easier

You seem to be looking for sort of confirmation bias towards how good AI is, if that is the case you should have just asked for it rather and seemingly asking for a discussion about AI and image editing.

Regarding sticking to technical inputs, Bharat's post does not do this yet you seem to find it "Wow" and "Interesting" but my post which actually did talk about AI and image editing is "thread hijacking"

Actually Bharat's post is a very instructive case - he wanted a clear result out of many variables and AI helped him find one. That was and still is worthy of a "wow !"

You said something about "skimming thru", I can see that is true. This is not a general discussion on AI and image editing - it is a discussion on the specific use of AI in Lightroom and maybe photoshop. The steps involved and maybe the issues if any.

My purpose of putting up this post was to share what I thought was an absolutely brilliant use of AI in everyday mundane post processing steps. To that end you could say that I am biased towards such usage of AI in Lightroom. No shame there. Do you know of a better usage or of any obvious disadvantage in getting 90% of the repetitive steps  done this way while using Lightroom or Photoshop? If yes, please share.

[g2]

This sort of lament is exactly what I was hoping to avoid. It seems one has only to say the word "AI" and all the Thread Hijack ghosts come out of the closet.  Let's stick to any technical inputs that you may have (if any) about how the man in the video made his bulk edits  easier

You seem to be looking for sort of confirmation bias towards how good AI is, if that is the case you should have just asked for it rather and seemingly asking for a discussion about AI and image editing.

Regarding sticking to technical inputs, Bharat's post does not do this yet you seem to find it "Wow" and "Interesting" but my post which actually did talk about AI and image editing is "thread hijacking"

[Krish Chandran]

AI is a force multiplier.
But you absolutely need to know what you're doing for it to give you the results you want.

Not necessarily.

I skimmed through the video briefly, and the fellow asked AI to analyze his work and then edit photos in his style.

What's stopping someone telling AI to analyze Ansel Adams work and edit their photos in Ansel's style? They can very well do this without any knowledge about the zone system or in depth processing.

All you need to know is how to give AI a prompt.

In a similar photographic context, a friend used AI to help create and fine tune a picture style that could emulate a Leica like rendering on a Sony camera.

Can he claim this picture style to be his creation, or call it his signature style considering AI helped create it.


There is a whole grey area around this.

Its a great tool for someone doing bulk processing (say a wedding photographer) but terrible for say a fine art photographer who spent years developing a  signature style, which can simply be replicated now with a prompt.

Studio Ghibli trend comes to mind as an example.

This sort of lament is exactly what I was hoping to avoid. It seems one has only to say the word "AI" and all the Thread Hijack ghosts come out of the closet.  Let's stick to any technical inputs that you may have (if any) about how the man in the video made his bulk edits  easier

[g2]

AI is a force multiplier.
But you absolutely need to know what you're doing for it to give you the results you want.

Not necessarily.

I skimmed through the video briefly, and the fellow asked AI to analyze his work and then edit photos in his style.

What's stopping someone telling AI to analyze Ansel Adams work and edit their photos in Ansel's style? They can very well do this without any knowledge about the zone system or in depth processing.

All you need to know is how to give AI a prompt.

In a similar photographic context, a friend used AI to help create and fine tune a picture style that could emulate a Leica like rendering on a Sony camera.

Can he claim this picture style to be his creation, or call it his signature style considering AI helped create it.


There is a whole grey area around this.

Its a great tool for someone doing bulk processing (say a wedding photographer) but terrible for say a fine art photographer who spent years developing a  signature style, which can simply be replicated now with a prompt.

Studio Ghibli trend comes to mind as an example.

[Krish Chandran]

Not exactly the same thing at all, but I was trying to mix a few paints (yellow, brown and white iirc) to get a particular shade.

Quite painful.

Made a few mix combinations, kept track of the proportions, painted adjacent patches on the test surface and then quite literally entertained myself watching paint dry.

Wasn't working. Too many variables, and not enough experience to juggle three colors to get the desired result.


Then the brainwave.

Uploaded a shade card, mentioned the three colors available, and asked for an appropriate mix.
Got a weird mix proportion that I probably wouldn't ever have guessed.

Result was spot on.


AI is a force multiplier.
But you absolutely need to know what you're doing for it to give you the results you want.



In a similar photographic context, a friend used AI to help create and fine tune a picture style that could emulate a Leica like rendering on a Sony camera.

Wow ! That is very interesting !

[Bharat Varma]

Not exactly the same thing at all, but I was trying to mix a few paints (yellow, red and white iirc) to get a particular shade.

Quite painful.

Made a few mix combinations, kept track of the proportions, painted adjacent patches on the test surface and then quite literally entertained myself watching paint dry.

Wasn't working. Too many variables, and not enough experience to juggle three colors to get the desired result.


Then the brainwave.

Uploaded a shade card, mentioned the three colors available, and asked for an appropriate mix.
Got a weird mix proportion that I probably wouldn't ever have guessed.

Result was spot on.


AI is a force multiplier.
But you absolutely need to know what you're doing for it to give you the results you want.



In a similar photographic context, a friend used AI to help create and fine tune a picture style that could emulate a Leica like rendering on a Sony camera.

[Thad E Ginathom]

avoiding unwanted rant except to say Utility for me? no-just-no

[Krish Chandran]

I Let ChatGPT Edit My Photos in Lightroom !! 

Stumbled across this fascinating video on YouTube  It demonstrates a massive shift in how AI interacts with post processing photography, rather than just doing standard pixel manipulation:

Why this is entirely different from typical AI photo editing that most us have been exposed to so far

• No pixels are being artificially generated or swapped out dynamically
• The AI actively operates your actual Lightroom software, mechanically adjusting the sliders and creating precise masks
• It preserves the full quality of your original RAW files exactly like a manual workflow would
• Every single setting remains completely adjustable, allowing you to fine-tune the final image 90% of the way through. You can still add to the edits to finalize the end result.

A quick request for the discussion:
I would love to get your perspective on the technical execution of this workflow and how it could potentially fit into our day-to-day processing.

Could we please keep the discussion strictly focused on the technology itself, its mechanics, and how we might use it to optimize our actual processing workflows?

 My request to those who wish to respond. We All  know the standard arguments regarding the "AI takeover" and the perceived loss of traditional post-processing skills, the downfall of mankind and so on. .  let's leave the philosophical  laments aside for this instance. That is not what this is about, so please do not hijack the thread.

 Let's look at this purely from a utility standpoint and a potential learning tool for the software we already use. How much we can use it in everyday functioning - who would need it most and so on. . 

Watch the video Here