Using AI to generate images

[Rick]

I've seen some here, Ron and TomO and others, post images of modeling ideas using AI.
I know I could watch YouTube videos to learn how to do it, but figured I'd get a head start getting input from members here.
Also, it might encourage others to try and do the same.

I know some AIs have free versions.
Can the free versions be used to generate images?
Can someone post the prompt they used to generate an image?

Thank you.

[deemery]

The only time I tried AI image generation was to add an eyepatch and a pirate hat to a picture of a friend's dog who had lost an eye.         Aaaarrrrrrrrfff!

dave

[Philip]

I told Gemini to remove the background on an existing photo and add water and forest to the background.

Rick, the best way is to just jump in. I have done a few prompts and it sometimes misses the mark.

[Philip]

Here is a prompt gone haywire!

 

[Larry C]

I may as well jump in here and add my "nickels worth" since pennies are no longer minted.

Philip Nice image of the boat in the water but the last one is interesting.

Ok there are some "free" AI's to use. Chat GPT has one, Gemini from Google also has a free version. Theses can be used through your browser. Gemini is too cartoonish for my tastes and Chat's works pretty well but you are limited to 5 images in a 24 hour period. Artspace.AI isn't bad for doing scenery stuff and offers it for $49 for life once in awhile.

Personally I use "Chat GPT Go" which is a paid version @ $8.45/month for doing Star Trek and I get enough use out of it to make it worth my while. If I ever want to get rid of it I can cancel anytime.

Rick yes the freebies will generate an image for you although the quality might not be the best. As far as a "prompt" is concerned most versions give you a box to type in and when finished you just click the button with your mouse. The trick is learning how to write the exact words in the box to get the image you want; also if it doesn't turn out the way you want you can always "edit" the image to get a better result. This is a case of "practice makes perfect; I've been fiddling with AI for the past year." Hope this helps and answers your questions.



Here's an image I edited from someone else, with permission, the other day showing the Primary Sickbay for a Starbase I write Trek on. The image desperately needed editing & cleanup.

[elwoodblues]

Rick,

I used ChatGPT to generate the the tunnel portal scenery.

The issue with the AI platforms is that there are usually limits to how many photo/ picture manipulations you can do per day (usually 3-5 per day) for the free versions.  I bought a subscription to Chat GPT for $25.00 per month.




This is the photo I started with.  I typed in the following in the AI box "This is a scene for a model railway.  Add mountainous scenery to complete the scene" (or something like that).

The picture it generated had a stone portal.  I then asked it to change the stone portal to a timber portal.





This is the final result.

Looking closely at TomO's attempts I see I need to refine my approach to generating these scenes.

[Larry C]

Here is the original model photo of Hank's Machine Shop:



And here's the version after using Chat GPT. The prompt was: "Around the image use a country town as a background". This is the result:

[Rick]

I used the free version of ChatGPT to create a photo realistic image of a single stall narrow gauge engine house in a wooded area.

I also wanted it to add a shop on one side but it only added a work bench.

[elwoodblues]

I don't know Rick, that looks pretty good.

[deemery]

The doors aren't wide enough to cover the opening...  The bench in front of the second door is strange, why 2 doors?  And what's that strange dark grey thing on the right side door?  Overall it looks OK, but when you look at details, you see "hallucinations".

dave