Same prompt with Midjourney 8.2 + ChatGPT + Ge

Hyper detailed portrait, beautiful Japanese woman, profile view facing right, head tilted back, eyes closed, lips slightly parted, serene expression. Flawless porcelain skin, soft red eyeshadow, deep vermilion lipstick, elegant elongated neck. Nihongami updo, glossy black hair with loose strands. Elaborate kanzashi on left: antique gold filigree, white chrysanthemum, crimson camellia, white blossoms, dark green leaves, turquoise/coral beads, crimson silk tassel, bira bira gold chain dangles. Single long red tassel earring with gold filigree cap. Layered silk kimono: ivory outer robe with gold leaf florals, indigo/teal peacock feather motifs, silver/gold embroidery; navy brocade inner with gold arabesque, white under collar, vermilion trim; crimson sash. Background: aged cream washi paper, mottled ink stains, gold leaf flecks, deep vermilion rising sun, gold cloud/wave drawings, white/black cranes with red crowns, red seal stamps. Ukiyo e meets hyperrealism, neo traditional Japanese, museum quality, dramatic soft studio lighting from right, cinematic color grading, antique ivory/gold/indigo/vermilion palette, 8K. No watermark, text, modern clothing, hanfu, frontal view, open eyes, extra hands, distorted face, blur. --ar 9:16 --raw --stylize 700 --hd

I think your observation is quite accurate — Midjourney does naturally tend to drift towards that polished editorial look, especially with female subjects.

But in my experience, that’s more its “default personality” than a limitation. Once you start playing seriously with parameters, style references, stylisation strength, image weighting, and very specific instructions for lighting and skin texture, you can push it quite far away from that typical MJ look.

That’s actually one of the things I enjoy about Midjourney: it has a very strong personality, so part of the fun is learning when to let it do its thing and when to fight against it.

For skin and lighting especially, I’ve found that precise control makes a huge difference. Sometimes I deliberately try to remove that polished magazine quality and introduce more imperfect, painterly or cinematic textures.

Thanks so much for taking the time to look closely and share your thoughts! 😊 I can definitely see why you chose the first one.

The first was generated with Midjourney, and I agree — I really like its texture and sense of depth. Of all four versions, I think it has the strongest spatial quality. The figure really seems to exist inside the image rather than simply sitting in front of a background.

The second was generated with ChatGPT. What attracts me most there is the colour palette — that aged, weathered gold. Once the whole image is unified within that restrained palette, it develops a kind of nostalgic decadence that I find very seductive. It feels as though there might be a story hidden behind it.

The third was Gemini. At first glance, it may actually be the most ordinary one. Some of the decorations and hair ornaments even look slightly artificial or cheap to me. But strangely, the longer I look at it, the more interesting it becomes. The geisha’s almost porcelain-like face, combined with the rest of the image, creates a slightly uncanny beauty that I didn’t expect at all. There’s something strangely haunting about it.

The fourth was generated with Qwen. That one feels relatively plain to me, so… I’ll politely leave it there. 😂

It’s quite interesting how four models can receive essentially the same artistic direction and still reveal four completely different visual personalities.