With AI graphic turbines evolving at these types of a rapid speed, folks have been placing them as a result of all kinds of intriguing exams. Someone’s applied a new feature in 1 of the most well-liked textual content-to-image equipment to see if it could be applied to redesign logos – and men and women even imagine some of the outcomes are superior than the originals.
Midjourney has opened up much more possibilities in its platform by incorporating one more know-how that has been evolving in parallel to image technology: picture description. One particular user has tried using working with it to explain legendary logos and then use the descriptions as prompts for new styles (see our piece on how to use DALL-E 2 for the basics of AI text-to-picture technology).
I questioned Midjourney v5 to ‘/describe’ some logos, to see how it would develop prompts for them, and to see what it would develop in reaction.Starbucks pic.twitter.com/EiKujhTl1wApril 4, 2023
In addition to employing ‘/imagine’ prompts to crank out visuals centered on a text description, Midjourney people can now do factors the other way all around, and reverse engineer photographs. By uploading an impression and making use of the ‘/describe’ prompt, they can question Midjourney to build a textual content description. That can then be tweaked and employed, to, in theory, crank out equivalent illustrations or photos with any variations that could possibly be sought after.
Because AI picture turbines really don’t create the exact image from the very same prompt twice, even utilizing the actual description attained through ‘/describe’ will produce versions of the unique impression. A person, who goes by the name of fofrAI on Twitter (opens in new tab), has tried out it out on logo designs, from Apple to Starbucks, Twitter and Pepsi (or must that be Pespi?), to see if it could ‘redesign’ their logos for them.
Really a large amount of persons look to assume the Starbucks brand redesign is far better than the authentic, whilst it is fussier and lacks equilibrium and symmetry. Some of the other patterns are just unusual. Since AI impression generators can not manage textual content, Pepsi’s even improved its title.
But I have uncertainties about the use case. The point that some of the logos produced are so equivalent to the genuine types seems probable to be a result of their inclusion in Midjourney’s coaching info relatively than the description by itself. Kind ‘Starbucks logo’ as a prompt into an AI graphic generator, and you will possibly discover that it is aware of what that is, even though the end result may glance horrendous.
And although AI impression turbines can be appealing instruments for throwing ideas all around, below any thriving thoughts for redesigns emerge by accident. Even adding unique descriptions to try to tweak the style and design in accordance to any strategies you may possibly by now have in intellect is nevertheless likely to be fairly haphazard for now, but we could shortly get to a point in which you can explain an current brand and tweak the description to make it less difficult or flatter (like we have witnessed so a lot of vehicle symbol rebrands) or to tweak the colors and other information.
Examine more:
More Stories
Jillian Warren ’24 Has Patterns for the Potential
What’s next for graphic design, according to industry insiders?
“A weird, wonderful and demanding job”: How Gina Alessi made the graphic design in Apple TV’s Manhunt