A/B tested Midjourney vs DALL-E for a week straight - one was way better for character art

So I had this project where I needed to generate a bunch of fantasy character portraits for a D&D campaign I'm running. I spent about 3 days trying to get consistent faces and outfits out of Midjourney v6, but every prompt felt like a dice roll. Then I switched to DALL-E 3 through ChatGPT for comparison. Honestly, the difference was huge. DALL-E just nailed the exact poses and expressions I described without me having to re-roll 50 times. Midjourney gave me cooler textures and lighting, but for specific characters with matching outfits across multiple images, DALL-E won easily. I think it comes down to how each model handles prompts - DALL-E seems to actually follow your instructions instead of interpreting them as vague suggestions. Has anyone else found one model way better for character consistency than the other?

2 comments

2 Comments

claire_ramirez2211d agoTop Commenter

Yeah, "every prompt felt like a dice roll" - that's exactly it. I see this same pattern everywhere, not just in AI art. People buy fancier tools thinking they'll get better results, but half the time the simple one that actually listens works way better. It's like when someone buys a high end espresso machine and can't make a decent cup, but a basic drip coffee maker just works every time. You nailed it.

sagey6911d ago

Is it really that deep though? I mean, we're talking about typing words into a box and seeing what pops out, not open heart surgery. If the fancy setup makes the person happy and they're not hurting anyone, does it matter if they could've used a free tool for the same result? Seems like a lot of energy to spend worrying about other people's coffee machines or AI prompts. But hey, if it helps someone feel better about their choices, more power to them.