I love midjourney for creative and artistic stuff but spent weeks trying to make it work for photorealistic images of the same person across multiple generations and finally gave up on that use case.
V6 is better but realistic human faces still have that slightly off quality on close inspection. And maintaining identity across different prompts? Forget it. Same character description gives me different people every single time no matter what seed tricks or character reference workarounds I try.
I'm testing foxy ai instead which takes the opposite approach, you upload reference photos and it trains on those to lock in the likeness. The realism is genuinely impressive and it lets you push into more suggestive territory which obviously performs well on social platforms. Way more limited creatively than midjourney but for generating consistent photorealistic content of the same person it actually delivers where mj kept falling short for me. I still use midjourney for concept art and mood boards and stuff where consistency doesn't matter, just accepted different tools for different jobs at this point.
Anyone else splitting their workflow across multiple tools like this or found something that handles both use cases?
[link] [comments] https://www.reddit.com/r/midjourney/comments/1r0stn2/midjourney_for_realistic_human_photos_feels/