Comparing the Top Free AI Image Generators
I did something unique this holiday season with my significant other. We had some time to burn on Christmas Eve and ended up spending it generating funny holiday themed AI images. I’m not sure it will become a holiday tradition, but it was definitely a novel experience. It also inspired me to compare the top free AI image generators (using mostly NYE theme prompts). First, let’s get started with how to use AI image generators for free.
How to create AI images for free
If you’re looking for a free generative AI image creator, I recommend checking out Bing Image Creator, Imagine with Meta AI and Adobe Firefly. Based on my research and testing these are the best free image generators available today.
Once you’re signed into your Microsoft account you can access Bing Image Creator via bing.com/images/create, chat.bing.com or Microsoft Copilot. Check out my previous article to learn what Bing Image Creator is and how to use it to get started creating images with DALL-E 3 for free.
Creating images using Imagine with Meta AI is really straight forward. First start logging into your Facebook, Instagram or Meta Account. Please note that you will need to create a Meta account if you don’t already have one associated with your Facebook/Instagram account. Next go to imagine.meta.com, enter your prompt and click Generate. Once your images have been generated you can share, download and/or report harmful content. Meta does include a disclaimer that images generated by AI may be inaccurate or inappropriate.
You can create AI generated images with Adobe Firefly by first creating an Adobe account. Next go to firefly.adobe.com and submit your prompt. It’s worth pointing out Adobe Firefly only provides 25 credits a month for free (each prompt uses a credit).
DALL-E 3 vs Meta AI vs Adobe Firefly
I used 5 identical prompts to compare the top 3 free AI image generators. Each prompt and the images generated are listed below:
And for the last comparison I decided to go with a more abstract prompt the wasn’t specifically NYE themed.
Here’s a few observations a made during this experiment:
- Meta & Adobe both add watermarks to images
- Although I didn’t measure the time it took each tool to generate an image, Meta AI seemed to be significantly faster across the board
- Meta AI & Adobe Firefly appear to struggle more with contextual details in prompts more than DALL-E 3
- Meta AI leans heavily toward photorealistic images
- Adobe Firefly offers a wide range of customization and control options that I didn’t use as part of this experiment, but appear very helpful
Issues with AI generated images
I noticed a number of issues with AI generated images while performing the comparisons above. Here’s some of the biggest issues I noticed with generative AI images:
- Text is often illegible and/or contain typos (even when included in the prompt)
- Some models really struggle with context (e.g. Imagine with Meta didn’t really understand the “nye ball drop as colorful paper cutout” prompt)
- Images containing people are often horrifying examples of the uncanny valley (specifically avoided in this comparison because it is a universal issue)
- Sometimes images bringing multiple subjects together (i.e. an old clock tower in large downtown city) results in images that appear poorly Photoshopped (e.g. the perspectives, proportion and lighting are inaccurate)
Other AI image generators
My comparison of free AI image generators isn’t exhaustive so I did want to call out some great alternatives:
Note that Canva (Google Imagen & DALL-E 3) – isn’t really a stand-alone image generator though because you have to do within Canva design (only 3 credits for free, at least with Google Imagen).
Why aren’t Midjourney or Google SGE included?
Midjourney discounted their free trials and require a paid subscription to generate images. Google supposedly allows users to generate images using search generative experience (SGE) however I haven’t been able to get consistent results. I was able to get SGE to generate some images (below), but I experienced too many challenges to include it as part of this comparison. In some cases I ran into the same issue as Barry Schwartz when attempting to test Google SGE’s generative imagery. In other instances, SGE simply stated “AI-generated images aren’t available for this search. Try creating something else.”