I like making artsy things, but have very little art talent or skill. The last couple of years I’ve leaned into AI image generators for my art references. The works below were largely created by Bing Image Creator, which is free, plus a small number using Ideogram.ai, which has a very limited free plan, but I spent $10 for a month just to see what it can create.
Everything was made using 2.5mm plywood, 5.0mm plywood, and 1/8" hardboard purchased from the local big box stores. The wall art measures 20" to 30" across the largest dimension, the coffin lanterns are 13" tall, and the cubes are 110mm on a side.
Yes, a Primo. It has a working area of 32" by 24". I’ve moved the size around a couple of times, but it is currently sized for 30 x 20 foamboard as well as cosplay props for my daughter. It is a bit big for quality routing.
Those look AWESOME!!! I have a primo that I only use for laser. Right now its 24" x 24" but I have been considering taking it out farther in X direction. Would be real nice to be able to set a full length strut on it and not have to do 2-3 set ups to engrave lol.
So ai does not care if it does licensed content hmmmmmm
I don’t think the law around AI content is settled. I read one opinion from the copyright office that indicated that AI content cannot be copyrighted. Copyright requires human creation. Anyway, I feel much better about potential copyright issues of AI created art than I do about the copyright of most found or purchased art on the net. On Etsy, you will find the same art being sold by multiple different people, plus you will often find that art being distributed on “public domain” websites.
What did you tell it to get the coffins?
So here is one of the prompt I used: “A photo of a laser-cut wood wall hanging in the shape of a coffin. Inside the coffin are cobwebs, bats, and the words “Trick or Treat”. The wall hanging has a black wood finish and is hanging on a white wall. The background is clean and simple.”
I made an ugly error when I created the coffins. I first designed and constructed the coffin shells, then I started creating the coffin fronts. The proportions of the coffins I designed and the proportions of the AI created art were different, so I had to finesse the graphics to get even frames around the content. I would have saved myself a bunch of work if I’d constructed the shells to match the proportions in the AI art.
Ahahahahaah, I meant just the lions, not the casket, Guess I need to learn to request to AI then all my real world communications will work better BAHAHAHAHAHA
Yeah that’s the free bing creator - I can’t really justify a paid sub to any of the LLMs but if they did a tier for under tenbux that included DALL.E I might be tempted.
Here are four from Ideagram. As mentioned above, Ideagram gives 15 free credits a day. I find Ideagram is superior when any form of text is included. Other engines tend to mangle or even greek the text.
Another I like to play with is ImageFX (Google Chrome). Most of the time I’m looking to generate silhouette images that are easily traced in Lightburn. ImageFX rarely produces this kind of image no matter what prompts I enter. What it does understand is multi-layer laser-cut projects. I’d have to recreate all the graphics, but it produces the vision for the completed project. Here are four from Image FX.
ImageFX is free. The daily limit seems to vary, but was over 100 generations (400 images) per day the last time I ran into the limit. Note you need to vary the prompt or the seed (manually) to produce a variety of results. The same prompt tends to produce the same images if the seed is not changed.
I’m using prompts that produce things for my laser. I could use ImageFX for inspiration for relief carving, I’d use “relief carved” in the prompt, and get things like this:
The art would still need to be converted/modeled to get an STL for relief carving.
If I was looking for something easy, I’d use prompts that tend to produce silhouettes. Silhouettes are easily traced to produce laser paths, router contour cuts, and pocketing. I’d get things like this (Bing and Ideagram):
For the four packs of pictures in my last couple of posts, I tried prompts several times, and then picked four I liked. All four did not come from the same generation. For me, generating AI images is a numbers game.
This evening I’m attempting to generate a silhouette of a wing for a project I’m working on. I’ve tried over 20 times (80 images), and I only like two of them for the project. To be clear, all 80 are wing silhouettes, but only two have a form that appeals to me. There are certain recipes, settings, and features that promote good images for laser cutting, but a lot is just tossing things at the engine to see what happens. I keep tabs open for ImageFX, Bing Image Creator, and Ideogram, and toss things at each periodically.
Crafting the prompt to get what you want is definitely a skill and it can be frustrating.
It’s not like google where you use a short phrase or keywords, literally describe in as much detail as you can exactly what you want. Example:
A piece of large art, a laser cut silhouette painted black on a white wall viewed from straight ahead. The art is an American football with the words “DETROIT LIONS’ on it below a lion.
Interesting that almost all of the footballs have white stripes. This is more of a college (and Canadian Football League) thing, while the NFL is stripe-free. Unless AI thinks the Lions have been demoted from the NFL???
