AI for Image Modification or Generation

trc65

Member
Full Member
Messages
7,790
Reaction score
24,362
Location
Cameron, Illinois
First name
Tim
On another turning forum there is a discussion in using AI. Evidently, there has been an instance or two of someone stealing an image of a turning, using AI to modify it and then presenting it as their own. Stealing has been going on since the Internet started, so nothing new there, but the ease of modifying images is taking this to a new level.

There are of course tools to help ID the use of AI, but one member ran a few through detection tools and found the same image was ranked by one tool as 90%+ modified by AI, and another tool predicted only ~50% chance of being modified by AI.

Anyway, the real reason I started this thread is they were also showing image that used AI to modify backgrounds with incredible results.

Another point to make is that most AI tools will imprint a watermark on the image, but as most know there are many tools available to remove watermarks.

I have never used any AI tool other than what shows up in an ordinary Google search, so was curious how easy they are to use.

In about one minute I went to Google, searched for "Gemini", went to the page, uploaded a picture, told it to change the background and saved the result.

I uploaded this photo:

PXL_20260428_041944519.webp

Told it to 'change the background to a white foreground gradient to storm gray background'. In about 30 seconds I got this image back:

1783468849797.webp

Quite a difference!

First thing to notice is the Gemini watermark in the lower right corner touching the one form. Second thing is it made a hard line on the shelf I'd the second row.

Will I start using AI, probably to play around with, especially to see what happens if you try to use it for color/white balance/contrast, etc. I will occasionally manually edit photos to correct those things if I'm having a hard time with a photo.

I may use it to play with backgrounds some, but when I do, plan to leave the watermark, and mention that I've used AI for the background.

So, what's the big deal.... Maybe nothing, but personally I think it is being a little dishonest to use AI in a photo and not mention it. For posting in this forum, no big deal either way, but imagine if someone is using this tech to represent items for sale, be it completed objects, or maybe pieces of highly figured wood for sale. I'm sure we've all seen images of impossible looking figure in wood that is posted in various places.

Comments, thoughts, further examples, all is welcome.
 
I really enjoyed the way you presented this topic Tim. Very thoughtful and insightful. So many folks I talk with just bash AI and aren’t objective. I also don’t like big media stories in the way they portray AI. Just my opinion but I don’t believe that AI is going to replace jobs in mass. If they do replace jobs in a certain area, I am going to look at those companies more closely and question them.

I had a conversation while I was on that fishing trip re this subject and the person knew a company that did lay off a bunch of folks. We discussed their lack of foresight and failing to see how to leverage AI to expand the company.

This contractor I am now working for is building an environment for the use of AI for the Air Force but the AF is taking the use to build maintenance models. They are ingesting huge amounts of data and then using the AI to dig, model, and report towards predicting failures, failure rates, etc. They like to say they have put “guard rails” on the prompts so the AI is very focused.

That got a little off topic gang, but what I am trying to get at is there are good uses. One thing for sure, I totally agree with Tim that a user should acknowledge its use whether helping an image or writing a description.
 
  • Thread Starter Thread Starter
  • #3
@Gdurfey

Interesting on the maintenance models for the AF. Is that for planes or another aspect? I ask as my nephew has spent his whole AF career at F.E.Warren working maintenance on missile silos and is now a Master Sergeant directing all those crews/teams. He is essentially a walking/talking AI model for that work, which is why he has never been transferred from that base.
 
I like doing my own stuff. I don't mind if you doing your stuff means AI gets involved. I do appreciate the acknowledgement of it's use. I feel that is only proper as we would do the same sort of thing if we got pro help from any other source.
 
I used ChatGPT to develop my itinerary for my 7-day long tourist trip to South Dakota last October. It was an amazing tool for that and will use it again for it. I've also used it for developing a purchase list and the cut list for installation of beadboard and trim at my son's house. Loved it
 
  • Thread Starter Thread Starter
  • #6
OK, this is some scary sh!t.

Took this image and told Gemini to change the form shape to a degikup bowl shape and change the background.

PXL_20260426_054808418~2.webp

This is the result. The shape isn't a degikup as that shape doesn't have a narrow neck and flared mouth. However, it is a pleasing shape. AI also took it upon itself to finish off the interior for me. This is just too freaky.....

1783486399630.webp
 
A very true representation. We should update the rules to explicitly allow the use of AI as long as the usage is stated in the same post.

In the world of woodworking, AI is not nearly as dangerous as in other areas (though the potential for scamming is greatly increased) I think it is still important to set clear rules so that abuse is limited. Thankfully this is a very civil and friendly forum and I don't think we will see much if any malicious use but it is still important to set standards.
 
I've started playing some with Gemini. Mostly with car pics. It is astonishing what it can do.
These two pics are both generated. I used a picture I took of the car earlier. It replaced the backgrounds and altered the position of the vehicle.
I will continue with it to learn more.
mymagshot.webp e51a1596-46fc-41bf-8277-d936d9bd679c.webp

Started with this one I took with my Iphone..
IMG_4464.webp
 
Last edited:
Like most things, AI was developed to be beneficial and helpful but a small per centage of the population finds a way to use for bad stuff - like making fake photos of someone doing something they shouldn't be doing, making something look better than it is so they can sell it for more, etc. I use Photoshop to correct some of my photos, crop off stuff I don't need, rotate a scenic shot so the horizon is level (I think my eyes are crooked sometimes), adjust the white balance on a bright sunny shot etc. But I essentially tell the program what to do and how much to do it using sliding adjustable scales or per centage change amounts, etc. I don't trust the computer to do stuff autonomously because it feels like I've lost control. Plus I don't trust that my image would then be stored somewhere else and is available to someone else without my knowledge. I don't understand it completely but in many situations I think AI makes people lazy and robs them of control of their own material. When it comes to photography, AI is going to make more suspicious than I already am.
 
@Gdurfey

Interesting on the maintenance models for the AF. Is that for planes or another aspect? I ask as my nephew has spent his whole AF career at F.E.Warren working maintenance on missile silos and is now a Master Sergeant directing all those crews/teams. He is essentially a walking/talking AI model for that work, which is why he has never been transferred from that base.
Tim, this is a new technology effort and currently focused on the supply chain data folks at Tinker. The first folks using it is the B-52 community. But it is rapidly expanding use including to the other 2 depot bases. Tinker AFB is OK City, the other 2 are Robins AFB, Robins, GA, and Hill AFB at Ogden, UT. Hill is responsible for missile maintenance. But all 3 of these depot organizations are looking at this technology. The future is great, just slow to roll out.

Hope folks didn’t mind the lesson in Air Force logistics.
 
AI can do some amazing stuff, but it bugs me that it is getting to the point where it is hard to determine what is real/true on the internet anymore...
 
I wanted to bring this discussion back to life. I have been playing around with ChatGPT quite a bit recently. I have a couple use cases that are worth talking about.

1. I used it as a coach to help me learn how to improve my photo setup and take better pictures. It did a great job helping me improve pictures without actually doing any editing to the photos. Low risk, high benefit. It helps me do my own work better.

1b. Editing photos has been a mixed bag. I have tried to get it to make adjustments I would make in photo editing software if I knew what I was doing. It keeps trying to modify the object. I have gotten better at giving specific instructions that control the output, but it takes extra iterations and careful checking the output. High risk, medium benefit. The potential is there, both for good and for bad.

2. I have taken a picture of a block of wood and asked for ideas of what to make with it. I got some interesting ideas and renders of what could be made. I have since then added some images of items I liked to build a library to help in the future. It is like an interactive pintrest. Low risk, moderate benefit. It helps me do my own work, but increases the influence on what I make.

3. I am playing around with pricing on items. I first took pictures and asked for a price and it was surprisingly in line with my thoughts. It got a lot better when I feed it some price sheets and results from shows I have done. Medium risk, moderate benefit. My pricing data could get out into the internet, but I don't mind sharing it anyways, it has helped me clarify my thoughts around pricing.

4. I have done a few club demos in the last few years. One demo is actually a mixed turning demo and conversation about selling my work. I am doing it again for another club next month and fed my outline into AI to get suggestions to improve the presentation. It has done a good job coaching me and helping me clarify my thoughts. A good piece of advice I have gotten from someone else is don't use AI for your first draft or your final draft. I think that shows up really strong in this use case. Medium risk, high benefit. It could be used to come up with unoriginal content, but it also has potential to refine my own work. I see this use case as "spellcheck on steroids".

What other use cases do you have?
 
The company I work for is huge into leveraging AI. The above is great and thanks for sharing. It’s all about the prompt and learning. My coworkers like to use the term “guardrails “….you have to put the guardrails up for a more accurate response. That is, more details in your prompt are the guardrails.
 
I work for Nike, but not designing shoes. But just for fun, I used AI today to design a shoe based on my Woodturning. I am planning to use this as a team activity to get people using AI because we are also pushing adoption here.

IMG_1153.webp
 
  • Thread Starter Thread Starter
  • #15
I haven't used it for image mods or generation since the beginning of this post, but the one thing I do use it for is as a better version of a search engine than what's been available since Google's early years.

I started using Google shortly after it became available and it was nothing short of miraculous at that time. I was a regional commercial veg production specialist with the Extension service, but also had to field questions from the public on pretty much any topic related to anything living. Entomology, plant pathology, weeds, general horticulture, lawns, landscaping, gardens, houseplants, pets, pests in the home, you name it and I'd get questions.

Back then, type a word or two into Google and usually on the first page of results would be one or more links to University fact sheets, or research trials. I used a phone headset to keep both hands free, and it got so I would have the information available before the client finished explaining their problem. I'd explain what they needed to know, print a fact sheets or two and transfer them to support staff so they could get an address. I'd print the fact sheet to the support staff desk, they'd print a label and have it postmarked in a minute or two.

Now, when I want to find something related to pesticide labels for certain problems (weed, disease or insect for eg.) it is painful using regular Google. It's difficult finding any info that is university generated, but rather is page after page of commercial listings trying to sell things mixed in with all the self proclaimed "experts" blogs that have plagiarized (often incorrectly) info from each other. It's like the old game "Telephone" or "Chinese Whispers". Copy some root info from a source, but change a word or two so it's not an exact copy. The next person down the line does the same, and pretty soon the info is incorrect, or at best incomplete enough to be misleading.

Google search will now give an AI overview, which gives a source for each paragraph. It's not perfect, but is much better than it has been for the last 20 years. You still have to cross check the info for accuracy, but it gets you on the right track much more quickly.
 
Back
Top