Summary

The author, a writer with limited art skills, reviews three free AI art tools—StarryAI, Wombo Dream, and DALL-E—to create illustrations for a story about dragons, ranking StarryAI first for consistency and usability despite some inaccuracies, followed by Wombo Dream for its diverse styles and quick generation, and DALL-E last due to its less satisfactory and often abstract results.

Abstract

In the quest to enhance a fiction story about dragons with appropriate illustrations, the author, who admits to having art skills that peaked in kindergarten, turned to AI-driven art generation tools. The article details the author's experience with three such tools, evaluating them based on the quality, style, and accuracy of the generated artwork, as well as ease of use and cost. StarryAI emerged as the top choice, offering consistent artwork and multiple options per prompt, though it required credits for extended use. Wombo Dream came in second, praised for its wide range of artistic styles and rapid image generation, but it necessitated significant trial and error to produce usable art. DALL-E, despite its reputation, delivered the least satisfactory results, often producing images that were too abstract or not fitting the prompts. The author notes that regardless of the tool used, additional image manipulation was often necessary to achieve the desired outcome.

Opinions

StarryAI is recognized for its ability to produce consistent artwork that aligns with the user's prompts, despite occasional inaccuracies in interpretation.
The author appreciates the variety of styles offered by Wombo Dream but points out the need for extensive iteration to obtain usable images.
DALL-E is criticized for generating images that are either too abstract or fail to accurately represent the prompts given.
The author expresses a preference for free tools and is reluctant to pay for premium features, such as different aspect ratios or the ability to iterate on an image.
The necessity of using additional image manipulation tools like GIMP or Canva is acknowledged, as AI-generated art often requires further editing to fit the author's needs.
The author concludes that while AI art tools can produce images in the vicinity of what is desired, the user must still refine these images to make them suitable for their purpose.

Using AI to Create Art for Your Stories

Top three free tools ranked

Created by Author using Wombo Dream, Canva, and GIMP

As a writer, I specialize in words, not pictures. My art skills peaked just after kindergarten. This caused a problem when I posted one of my fiction stories about a dragon.

Because it was a fairly long read (around 4500 words, or 30 of my Daily Cuppa short-form articles strung together), I decided to break it up with illustrations. But I wasn’t going to find photos of dragons using computers on Pexels or Unsplash. I was going to have to create those illustrations somehow.

The new crop of AI text-to-art tools came to my rescue. Of the three I used, here is how they rank from first to last. I give examples of the sort of output you can get from them. Also, I explain how easy they are to use, and what I liked and disliked about them.

#1 — StarryAI

The best tool for my purposes was StarryAI. It created consistent artwork in the same style for every prompt. (You select the style or styles you want.) It also gave 4 different options for each prompt, which was good, since it wasn’t very accurate in depicting what I was asking for. Usually, I could make one of the provided illustrations work. The downside is that they cost credits. You start with 5 free credits, but I didn’t want to pay for more.

The above graphic shows how the AI doesn’t quite understand what you’re asking for. I asked for an image of a fat man holding a fat baby dragon. It gave me a fat man holding a fat baby, both of whom are at least somewhat dragonish. (I assume the dark grey wing is meant to be coming out of the back of the man’s head, not part of an otherwise invisible dragon standing behind him.)

When I asked for an image of four dragons hatching from eggs in the sand, watched by humans, I got egg-shaped humans, surrounded by dragons. The art I eventually used had four dragons watching four eggs. I just clipped off the part with the eggs, and used the part with the dragons.

You can specify what aspect ratio you want for the artwork, but only if you’re a premium member. The free version gives you a portrait format. If you want different aspect ratios, you’ll need to use an additional image manipulation tool, such as GIMP or Canva.

Also, it works on one prompt at a time. You’ll have to let it complete what it’s doing for you before you can ask it to do something else. Fire it off, then come back and check on it in fifteen minutes or so.

#2 — Dream

My second-choice tool was Dream by Wombo. Artwork was in every style imaginable, and it took a lot of tries to get something usable. Even then, I had to be willing to do some image manipulation to get exactly what I wanted. You could iterate on an image you liked that was close, but only if you wanted to pay for it. I didn’t.

Here are some of the extremely dissimilar pieces that Dream came up with for “cute dragon working on a computer”. The first is the source image that became the title graphic for this article, with a little help from Canva and GIMP. And these are only the ones that I thought might possibly be usable. There were 30 or 40 others that weren’t even close.

All created by author using Wombo Dream.

As you can see, the four pieces I included were a fairly standard style of fantasy art, a photorealistic piece, and two paintings of various levels of abstraction. That fourth one looks kind of like it was based off of how Picasso might have painted a deconstructed dragon — all the right features, in all the wrong places.

The art is generated in seconds, not minutes. Which is good, because you’ll have to generate a lot of it to get anything you can use.

#3 — DALL-E

The worst tool, surprisingly, was DALL-E. My husband had an account, so he used a couple of credits to try and create a cute dragon working on a computer for me. The AI created photorealistic images that looked like plastic dragon toys sitting on keyboards. Or abstract images that hurt my eyes to look at, let alone try and figure out what was supposed to be the dragon and what was supposed to be the computer.

There is a free version of DALL-E. It used to be called DALL-E Mini, but is now called Craiyon. Judge for yourself regarding the results.

This AI was clearly the best at understanding what I was looking for — a cute dragon that was working on a computer. But the art skills are…not good. The images are just subtly wrong, like the laptop keyboard that is offset from the screen. The franken-dragon that seems to have been stitched together from pieces of other dragons. Or the computer screen that appears to be folded.

Conclusion

If you have a concept you want illustrated that you don’t think you can find stock photos for, you can try having an AI text-to-art program create illustrations for you.

I tried three free AI art programs, and rated them on ease of use and applicability to my needs. Whatever program you use to create the art, you’re probably going to need to do additional image manipulation in another program, such as Canva or GIMP.

The AI artists are roughly comparable to the AI article writers. They produce something that is in the ballpark of what you wanted. But it’s up to you to craft it into something you can actually use.

Jennifer

I write articles on self-improvement, writing, positivity, and living in the Caribbean. If you enjoyed this article and you’re not already a Medium member, please consider joining ($5/month for all you can read) with my referral link.