My preliminary enthusiasm for AI picture turbines wilted considerably after I found not one in every of them will generate repeat characters for a comic book or kids’s guide. All obtainable picture turbines will generate single photographs solely, with no strategy to get the identical character doing various things.
OpenAI’s Dall-E 3 is reported as fixing that drawback. Dall-E 3 is free with Bing Image Creator. My fast check of that declare was removed from passable. I requested the AI to supply a comic book of a boy saving his canine from drowning with textual content to swimsuit. Bing Creator produces 4 photographs for every immediate. Below are the three I obtained. It appears the AI figured solely three photographs have been wanted to convey the entire story of my immediate.
The prime two have the identical boy in each footage and is passable for a comic book. The backside picture has totally different boys, and one boy strolling on the water. The textual content within the second is unintelligible. Bing Creator will typically give defective footage from fast prompts. People with three legs et al. The drawback is perhaps with my temporary immediate. I’ve but to delve deeper into that. @artichokes_art has posted on twitter a system he claims allows him to create 99.9% constant characters with Dall-E 3. I’ve not acquired to testing the system but. I give his full publish right here for anyone to experiment with……
How to create 99.9% CONSISTENT CHARACTERS with DALL-E 3 I’ve examined dozens of tips for consistency!
Sadly, most are simply NOT APPROPRIATE
Mismatched shapes/shades gained’t impress a possible consumer
But THIS TECHNIQUE provides near-perfect consistency
From monochrome illustrations
to superior cinematic photographs!
Don’t overlook to bookmark this; no one has shared this system but! _____ To begin, let’s settle for the two rules for a sensible strategy to character design. 1. We goal for 100%…
