How To Get The Results You Want From Your AI Art Prompts
AI Art is everywhere. It’s artwork generated by a machine learning process. An algorithm has “learned” some information, and used it to generate a new image. Photographers (and other artists) input a text prompt that sends the AI off to work and the results can be pretty stunning.
I am mixing AI Art with my toy photography. Because toys are meant to inspire imagination and aren’t constrained by reality, AI Art is a perfect marriage for toy photography. But it — like anything else, takes practice to get the results you want.
My first experiments with AI art went pretty bad. I thought I could just type “outer space” and I would get what I wanted.
The problem wasn’t that my AI software wasn’t working. It was in my expectations regarding the results I would get when compared with the effort I put into getting it.
It takes a little work, some vision, some practice and some experience to get what you want.
You need to figure out what text prompts you want to use to get the AI working in your direction.
The goal of this article is to teach you a workflow that will help you consistently generate a series of prompts that increase your likelihood of success.
Why is this important? Just typing in “outer space” gives a result like the one you see below…
The first step to getting the result you hope for is to really refine your thinking (and how you express it) starting with your SUBJECT.
So here the subject I used was OUTER SPACE. But that’s a pretty big and general term. I need to think about what it is in OUTER SPACE I am interested in. Is it NEBULA? Is it PLANETS WITH THREE MOONS? Is it STAR FIELD?
When you decide what your SUBJECT is and you carefully think about and express that, you get better results…
So I decided to refine my original text prompt and instead of OUTER SPACE, I typed — PLANET, SPACE, MOON, STARS.
That could be four subjects or in this case just one.
Looking at my new results — still not great.
I need to further refine my text input so next I want to think about MOOD. What mood do I want to set with this image?
I decided I wanted a GLOOMY mood — and when I added that to my original string of PLANET, SPACE, MOON, STARS I got several results to pick from that are starting to get closer to what I see in my mind’s eye but I am still way off the mark.
Since I am mixing AI art and photography, the most logical step is to use the MEDIUM I am working in to help further refine my AI result.
Mind you, I could be working in watercolor, oils, chalk, etc. but I am picking PHOTOGRAPHY — so let’s add that to the text prompt and see what we get…
This looks more like a photograph (which is what I wanted) so I am feeling better about my efforts. I still have work to do.
Next I want to think about my INSPIRATION for the piece I am creating. As a photographer, I love Ansel Adams. So I could use ANSEL ADAMS and add that to the text string to get the following result…
Just to have fun, and because I love juxtaposition in art, I will swap NORMAN ROCKWELL into the text prompt instead of Ansel Adams and this is what I get…
Fun, but not what I am looking for.
Next on the list of things to think about would be STYLE. Think about things like Art Deco, Anime, Pop Art, Postmodern, etc. I’ll try SURREALISM and add it to my text prompt and I get this result…
Pretty cool. I could stop here and be happy but I always like to go into this process with a checklist that helps me make sure I’ve really thought about what I want to make.
At this point I’ll recap where we’re at and what I consider to be a good AI creation workflow…
- SUBJECT
- MOOD
- MEDIUM
- INSPIRATION
- STYLE
But wait — there’s more!
Next up — I just want to influence the AI to give me an image that I can use as a background prop for my toy photography. COLOR is important because I might want to choose a complimentary color to get the best result. I will add the color PURPLE to my text prompt because I am using a toy that is a mix between yellow and gold — both of which go well with purple so here’s what I end up with…
I like this one. But there’s still at least one more step to consider. Add KEYWORDS that will help the AI have a better chance of giving you what you’re looking for.
Generally — you might think that the more information you offer here, the better off you will be. That’s often not the case. I suggest it might be time to try to refine your choice to one or two keywords. If you pick the absolute best keyword(s) you will get the best result.
I am simply going to add the keywords: HIGH RESOLUTION. I get a more photo-realistic result this way and if you look at all the iterations I’ve gone through, you can see how drastically each step has altered the final image.
HIGH RESOLUTION as the keywords gives me…
I like this. I can (and will) take this into Topaz Gigapixel AI (https://bit.ly/topazgigapixelAI) to upscale the image to match the resolution I am achieving with the photographs I will composite onto the AI background. Then I might take this into Photoshop and/or Boris FX Optics to further tweak the background to give me something truly unique and useful.
CONCLUSION
I Am Using the following workflow:
- SUBJECT
- MOOD
- MEDIUM
- INSPIRATION
- STYLE
- COLOR
- KEYWORDS
I can write better prompts that give me the results I am lookin for more quickly.
I will note that you can choose to eliminate any or all of these steps (except for subject) and you might get what you want. Maybe color isn’t important to you so if that’s the case, just don’t add a color prompt. If you don’t have any particular artist’s work in mind as inspiration, leave it blank. These suggestions are just designed to give you a framework that you can use to get started.
This is an area that is rapidly evolving and I suspect that we will get a more visual interface from some company (and soon) that makes this process easier. But for now, I suggest starting with my workflow as you think through your text prompts and wish you good luck with your AI art creation.
For a list of my toy photo gear and props go to:
bit.ly/toyphotogear
Follow me on VERO
vero.co/scottbourne
Roughly 90% of my photos are finished in Topaz Labs’ product called Topaz Photo AI. It is a one click sharpening, noise reduction, resolution enhancement tool that uses AI to see what your photo needs and only applies as much correction as is required and only to the areas that require it. Check it out at bit.ly/TopazLabsPhotoAI
