Preface
We have seven dAIs of Christmas to go, which in the song consist of a series of “jabs”—repeats of the same five-note ditty leading up to the “uppercut” of yesterday’s climactic fistful of gold.
Given that observation and guessing that most of you are in the midst of a busy period, what with holiday preparations and end-of-year business responsibilities, I’m going to dash through a series of lists of favorites to close out the 12 dAIs series. After today’s exposition on DALL-E, which I’ve used extensively and have received many questions about, there won’t be much to read, but I do hope there will be some good content to inspire bookmarks for exploring in the new year!
🎶 On this sixth day of Christmas, StrefaTECH explores … Six Image Generators 🎶
In this StrefaTECH series, I’ve been striving for a balance among many themes, including “modality,” which is the term used in AI to refer to the input or output type—text, image, audio, or video. Each issue includes generated images—a cover image plus one or more graphics to punctuate the article—most of which are AI-generated.
To use any of the tools in today’s list (below), you start with a prompt, much like with a chatbot such as ChatGPT. The state-of-the-art in image generation is notably behind text generation, though, when it comes to the AI tool providing a response that aligns with your instructions. And analogous to the hallucinations in AI-generated text, there often are various oddities in AI-generated images, some of which have been highlighted in prior StrefaTECH articles.
That said, what’s possible today is incredible! For free, you can generate high-quality, creative imagery within minutes in a very wide array of styles. If you haven’t yet tried it, do so now—it’s fun!
But as with searching for “just the right photo” on a stock image library site, be forewarned—creating “just the right image” using AI tools can be a shocking time sink. I’ve found that I may spend a third of the time taken to create a StrefaTECH post just fiddling around with images.
With that quick setup, here are six image generators worth knowing about. There are lots more, but these should give you a good foundation for knowing what’s available!
1. DALL-E 3 with ChatGPT 4 or through Microsoft Bing/Copilot
As is evident from StrefaTECH image credits, this has been my go-to image generator. If you need a good image, but not necessarily a professional-grade one, DALL-E 3 is tops.
DALL-E 3 is the image generator for ChatGPT 4 (the paid subscription) and Microsoft Bing (being rebranded as Copilot).
To access it through either chatbot, you simply include in your prompt that you want an image; e.g., “Generate an image of a professional woman standing on a soapbox speaking into a megaphone. Setting is a computer datacenter. Style photorealistic.”
Both ChatGPT and Bing will then call on DALL-E to create a unique rendering. As with a text-based conversation, you can prompt ChatGPT for changes, and it will retain the prior context.
Here’s how they work.
DALL-E 3 in ChatGPT 4
ChatGPT will usually elaborate on your prompt to add details that might elicit a better image. If you want to know the actual prompt that generated the image, ask for it. In this example, I followed up by prompting, ”What exact prompt did you use to generate that image?” to which ChatGPT responded, “A professional woman, wearing business attire, stands on a soapbox in a computer datacenter. She is speaking into a megaphone, surrounded by rows of server racks and data processing equipment. The setting is modern and technological, with a focus on the contrast between the human element and the high-tech environment. The style is photorealistic, capturing the intricate details of the datacenter and the determined expression of the woman as she communicates passionately”
Note that this is the only image generator on this list that requires a subscription (the $20/month ChatGPT Plus); the remaining five all are available for free.
DALL-E 3 in Bing
Bing responds to your exact prompt (without the elaboration of ChatGPT), but an advantage to Bing is that it generates a grid of four images. You can click on an individual image to see a larger version and can download it from there. You also can refer to them by number in follow-up prompts; for example, you might respond, “Generate more like 4. Replace the stilettos with flats and don’t have her look so angry.”
As a note, ChatGPT previously also generated four images, but sometime earlier this fall regressed to generating a single image. This has prompted me to use Bing much more often for image generation—I often want to see creative variations without having to ask for more options!
2. Canva Magic Image
One of the trends in AI that’s been increasing and is expected to continue into the future is the move away from the general-purpose chatbots/tools (e.g., ChatGPT) toward AI capabilities being embedded in the software we already use. Image generation is an early example of this, where graphic design tools that work with images are incorporating the ability to use AI to generate fresh images in addition to their traditional libraries/upload options.
Canva and Adobe are two of the most-used general design tools by nonprofit organizations, and both have had AI image generation for many months.
Canva’s image generation is part of its “Magic Studio” suite (Magic Erase, Magic Grab, etc.); it’s one of the free AI features and can be accessed through various elements. For more info on getting to the AI image generation, see this article.
At first (a scant few months ago!), I found the Canva images to be rather lackluster. But today, they are often at par with what I get from DALL-E, and if I’m already in Canva, I appreciate accessing the AI generation without switching to another app, exporting, and importing. Canva also has a selector for various styles, which I find to be helpful. So if you’re a Canva user, I recommend giving it a try!
3. Adobe Suite
Adobe has created its own generative AI tool, Firefly, which is embedded in its suite of creative tools. Adobe Creative Cloud, which is the low-end tool that most directly compares with Canva, incorporates image generation much like Canva.
I have very limited experience with Adobe, so rather than trying to contrast with other tools, I’m going to highlight that the capabilities exist and allow you who are Adobe experts to experiment with what it can do!
4. Midjourney
For most of this year and still today, Midjourney is benchmarked as the AI image generator that produces the highest-quality output. Unfortunately, it’s still only available via a somewhat awkward interaction in the Discord app, though they have recently announced limited access to a new more user-friendly interface that should make it much more accessible. And because of high demand, it seems that the ability to generate images for free has been suspended, so I’m going to put a pin in describing Midjourney for another time!
5. Stable Diffusion
Also on the list of high-quality image generation is Stable Diffusion. It powers many image generators in other tools, including in Canva mentioned above. It’s also available as an image generator, much like DALL-E, through the SDXL site.
6. Krea.ai
I just ran across a fairly new image-generation app that I’m adding only because it’s very different from the others and may represent a new breed of image-generation tools. It’s Krea.ai, and rather than describing it to you, I’ll just point you to this video demo. Give it a try!
Conclusion: “A Picture is Worth 1000 Words”
Generative AI isn’t just about conversing with a chatbot—generating cartoons, sketches, artwork, and photos is now at your fingertips. Think about ways to step up what you’re doing by adding a graphic. After a friend of mine first played with DALL-E, for example, she added a graphic to a simple event flyer for her neighborhood association and received wide accolades!
And as always, remember that if you’re going to use the results of your work with AI for anything going out to the public…
Make Good Choices