If there’s one thing generative AI has struggled with so far, it’s words. Ask a service such as Midjourney to create a new company logo and you’ll likely end up with a typographic jumble. Ideogram – a new AI service founded by a group of former Google researchers – has a distinct flair for typography.
Whether it’s a company logo, a slogan or a social media post that you want generated, Ideogram is the best AI service we’ve seen yet at combining words and images.
The service is capable of creative typography, such as spelling out the word ‘fish’ in the shape of a fish:
It can be used to create designs for t-shirts or social media:
It also has a flair for those twee motivational messages that you’ll find clogging up services such as Facebook and LinkedIn. It created the image below from the following prompt: A pencil transforms into a rocket, signifying taking simple ideas to new heights, with caption “new heights”, cinematic, photo, typography:
All of the images above have been created by users on the site, they’re not sample or publicity images.
How does Ideogram compare to Midjourney?
Midjourney has been our favourite generative AI service to date, but there’s no doubt it struggles when it comes to putting text in images. We fed the following prompt into both services: Modern beer can design for a beer called Dark Angel.
This was the best image Midjourney returned:
The design is strong, especially with the wing graphics in the background, but the wording is typically jumbled.
Ideogram returned this:
The design is less impressive, but the keywords are correct, even if the slogan at the top of the can is complete gibberish.
Perhaps a fairer comparison is to Adobe’s Firefly service, which has long-supported text effects. Here you can enter the words you want illustrated and the type of text effect you’re looking for. Here we asked for the words ‘Sweet Factory’ to be given a ‘candy’ effect:
Ideogram is currently free to use and there doesn’t appear to be any limit on the number of images you can generate. However, downloaded images are currently low-res (1,024 x 1,024 for square format images), meaning you’ll likely need some kind of upscaling to use the results on anything but the web.
More articles about generative AI art services
Nathalie Parent, Chief People Officer at Shift Technology: “HR is the conscience of an organisation”
For more than 30 years, Nathalie Parent has led global HR teams, working primarily with software companies. Today she’s Chief People Officer at Shift Technology
Amazon introduces new storage class that makes it cheaper to store rarely used files
Robot carers are real, but caregiving has bigger problems, writes Richard Trenholm in this FlashForward edition