Playing with AI

[OKW Lover]

I got caught up with the rush to everything AI. Decided to play a bit with some of my pictures just to see what it could do.
Using ChatGPT I gave it this image:



and this instruction:
Make this image look like it was taken at night. Add a spotlight effect
Which gave me this result:



Anybody else trying significant edits using AI? What product do you use?

 

[klmall]

I enjoy seeing some of what is created for fun using AI as done above but I am such a realist that I can find it confusing and disappointing especially when not initially stated as such. I have occasionally used AI in Photoshop and Topaz products but try not to rely on them. It's an exciting time and I wonder what lies next!

 

[hardcorestitch]

AI better not replace real photography *sighs*

 

[OKW Lover]

Anther experiment. I started with this picture
And told ChatGPT "Make this image look like it was taken at night. There should be lights in the windows and a crescent moon in the sky"

The result

 

[OKW Lover]

Also seeing what AI can do with this picture:

I asked ChatGPT "Use this picture as the basis for a highly detailed hdr image": and got this result:

 

[OKW Lover]

Changed this picture to an HDR version:

 

[jec6613]

I use effectively zero AI, at least in the common parlance used nowadays with LLMs and generative AI - things like turning something into, "HDR," involves inventing detail that wasn't there based on a prediction algorithm, and isn't actually high dynamic range. In both of the examples above you can see detail in the generative AI that isn't there in the original image.

I do use machine learning fairly extensively to de-noise some very high ISO, Kodak's ICE technology, and obviously ML algorithms make autofocus basically magic nowadays, but we've had earlier versions of all of that since the 1980s.

Below is an example of a single image I took a couple of weeks ago with more dynamic range than even the two examples above had in either their original or generative versions, and the camera still has several times more dynamic range with real detail in the raw file I generated this from. Nothing AI, just look at it at 100% and you can see the grit in the concrete.


PS: serious bonus points if you can name where this is from. And, no, it's not Disney related, just a picture I had to hand.

 

[anrew]

i tried to create a room design, but he didn't understand simple instructions

 

[klmall]

I'm sure others here have already done this. For the past few weeks I have been using Google's AI (Gemini models) to annotate my Disney and other trip photos with some interesting results. I upload a photo and see what turns up as a result for location or for more specifics on what something is. Often I ask for more than the AI Overview that comes first by going to AI Mode (which has 2 different Gemini models to use if needed) and I often ask for more specifics in the "Ask anything" section at the bottom such as "What store is this sold in the Japan Pavilion?" or "When did this close?" etc.

Now for the results! Sometimes it is right on and verifies what I thought or have found elsewhere. For places/things in the World Showcase it sometimes spits out generic info about a building style or location or historical reference but then I use the "Ask anything" option to narrow it down to a specific pavilion or place etc. Nine times out of ten that gives me the exact info I want. I do read the 'sources' listed for different AI statements and of course, determine how valid they are.

It really helped me with my Galaxy's Edge photos since while I'm a fan I know absolutely zilch about some things like specific ride characters and starships etc. Some of the 'sources' for this category are obviously fan-based and a little more suspect I think.

I do find contradictory information sometimes and outright errors but these are rarer than I expected. Google has a process to give Feedback which I did use once to report an obvious error in the naming of an attraction but I thought it was cumbersome and probably will never use it again at least with my WDW photos.

For non-Disney trip photos as expected I found it accurate for popular places like famous mountains and places and names of flowers and trees. But get off the beaten path a little and it gives generic descriptions and offers several suggestions for the location which may or may not be helpful. Even when it cannot give the exact location info you want it often still gives related facts.

DH and I have also used it to track down where to find specific items for the bathroom remodeling we are just starting. If we see a tile or whatever we like we use a picture of it on the computer to get more info. I know that you can do the same things on your phone etc. but for us as 'seniors' it is easier.

I'm sure that I am way behind in using Google's AI to do what I have described and I have read about the risks involved with using AI. Has anyone else here been using Google's AI?