Most base models flatten “Asian female” into a blended Western-Asian look. The 12 prompts below push the model back toward authentic East Asian features by naming the country, the eye structure (monolid / double eyelid), and the skin (glass skin / warm undertone). They cover the mainstream Chinese, Japanese, and Korean directions and are written to run as-is in Midjourney v7 and Flux 2.
TL;DR: Six fields fix the default — country, eye structure, skin, hair, restrained makeup, and a cultural scene. Use the templates verbatim, then swap the wardrobe and location to fit your brief. For a repeatable face across shots, lock identity with Midjourney v7 Omni Reference or InstantID on the SD side; the prompt controls the scene, not the person.
What a high-quality prompt should contain
Six fields are enough to hold a stable East Asian identity:
- Country explicit: write
Chinese / Korean / Japanese, never justAsian - Eye structure:
soft monolid eyes/neat double eyelid/slight epicanthic fold - Skin specifics:
fair skin with warm undertones/glass skin with subtle highlight on cheekbones - Hair:
long straight black hair/ash-brown shoulder-length/glossy short bob, not “beautiful hair” - Restrained makeup:
light pink lip balm only,glossy nude lip, no other makeup; heavy makeup pulls the face Westward - Cultural scene: Beijing hutong, Suzhou garden, Seoul side street, Tokyo cafe; the location locks the tone further
12 copy-ready prompt templates
1. Girl-next-door Chinese
Best for: Xiaohongshu / short-video lifestyle
A 24-year-old Chinese woman, soft monolid eyes, fair skin with subtle warm undertones, natural skin texture and visible pores, long straight black hair, light pink lip balm only, gentle slight smile, soft north-facing window light, ivory linen blouse, 50mm f/1.8 Fuji color, --ar 4:52. K-beauty glass-skin editorial
Best for: Beauty / fashion brand K-aesthetic
A 26-year-old Korean female model, sharp jawline with soft cheeks, glass skin texture with very subtle highlight on cheekbones, glossy nude lip, neat double eyelid, jet black sleek hair down, oversized cream blazer, neutral cool studio backdrop, single soft top light, 85mm f/2.8 K-beauty editorial, --ar 4:53. Tokyo cafe documentary
Best for: Japanese lifestyle brand content
A 23-year-old Japanese woman with a slight wave to her shoulder-length black hair, light freckles across the nose, soft monolid eyes, very natural skin texture, beige knit cardigan, sitting in a soft afternoon-lit Tokyo coffee shop, blurred warm wood interior behind, 35mm f/2.0, documentary feel, --ar 4:54. Modern Chinese courtyard
Best for: Tea / incense / Chinese-brand campaign
A young Chinese woman in soft floor-length linen dress, walking slowly through a sunlit traditional Chinese courtyard with a single old plum blossom tree, hand brushing past a blossom branch, soft afternoon golden light, modern minimal interpretation of Chinese aesthetic, 50mm f/1.8 Fuji, --ar 4:55. HK-noir bar editorial
Best for: High-end brand, cinematic personal portrait
A 25-year-old East Asian woman in a sleek black backless dress sitting at a luxury hotel bar at night, single warm tungsten side light, deep moody contrast, glossy hair, slight smirk over shoulder, anamorphic 50mm Wong Kar-wai red-amber palette, --ar 4:56. Suzhou-garden modern hanfu
Best for: Modern Chinese-style brand, content cover
A 27-year-old Chinese woman in soft white silk hanfu (modernized cut), sitting on a stone bench by a koi pond in a Suzhou-style garden, slight breeze in sleeves, looking down at her reflection, soft overcast daylight, 85mm f/2.0 cinematic, --ar 2:37. Seoul street candid walk
Best for: K-lifestyle brand, street photography
A young Korean woman in oversized white shirt and high-waist trousers walking down a sunny Seoul side street, slight motion blur, candid expression, ash-brown shoulder-length hair, soft 4pm golden light reflecting off cream buildings, 35mm f/2.0 documentary Fuji color, --ar 4:58. Tatami documentary
Best for: Japanese brand, lifestyle content
A 24-year-old Japanese woman lying on tatami in a sunlit traditional room, soft east-window light striping through shoji screen, hair fanned around her, eyes closed, plain white yukata, calm dreamy mood, 35mm f/2.0 documentary, --ar 1:19. Minimal modern qipao
Best for: Contemporary Chinese-style brand, designer label
A Chinese woman in modern minimalist silk qipao (knee-length, simple cut, no embroidery), standing in front of a plain stone wall under a single overhead skylight, neutral expression, glossy short bob, 85mm f/4 Hasselblad medium format editorial, --ar 4:510. Apartment window candid
Best for: Home / interior brand, personal portrait
A young Chinese woman in oversized fisherman knit and beige slacks sitting on a worn wooden chair near a sunny apartment window, glass of water on the floor, candid laugh at someone off-frame, soft natural window light, 35mm f/1.8 documentary, --ar 4:511. Post-rain wet-hair still
Best for: Cinematic portrait, brand narrative campaign
A 26-year-old East Asian woman, wet hair from rain pulled back, droplets on cheekbones, neutral expression, plain black turtleneck, single warm key light from camera-left, deep moody dark background, anamorphic 50mm cinematic, --ar 4:512. Dried-bouquet warm still
Best for: Tea / herbal / nature-brand content
A young Chinese woman holding a bouquet of dried Chinese herbs and dried sunflowers, standing in front of a soft taupe wall, looking gently down at the bouquet, oversized linen smock, neutral expression, soft overcast natural light, 50mm f/2.0, --ar 4:5Which model to run these in (June 2026)
The --ar flag is Midjourney syntax. The body of each prompt is model-agnostic, so the same text transfers to Flux and Stable Diffusion; only the aspect-ratio control changes.
| Goal | Best tool (June 2026) | How to set it |
|---|---|---|
| Most photorealistic skin (pores, undertone) | Flux 2 (Black Forest Labs, shipped Nov 2025) | Drop --ar 4:5; set aspect ratio in the UI, keep the body text |
| Most consistent face across scenes | Midjourney v7 | Omni Reference tab, strength 300-500 |
| Local / open-source, full control | Stable Diffusion XL + InstantID | Single front-facing reference image |
For aspect ratio, 4:5 is the standard vertical social crop, 2:3 reads as classic editorial, and 1:1 suits avatars and grids. A short primer on these ratios lives in AI Image Aspect Ratio Guide.
Common mistakes
Asian womanalone defaults to a Southeast-Asian plus Western-aesthetic blendbeautiful Asian girltriggers an Orientalist stereotypekawaiiorcute anime styleswings hard toward anime and kills realism- Mixing Chinese / Japanese / Korean cues in one prompt; the cultural markers cancel each other
big eyesis a Western-beauty keyword that erases authentic East Asian eye shape
How to push results further
Chinese facial featuresis more precise thanChinese woman(anatomy vs. identity)- For K-beauty glass skin:
glass skin texture with subtle highlight on cheekbones, dewy not oily - Japanese naturalism:
natural skin texture with light freckles,barely-there makeup - Modern Chinese aesthetic: drop the period words; use
modern minimal Chinese aestheticplus a specific scene - Asian-actress film-still: combine film-still grammar, country, and a Wong Kar-wai color reference
FAQ
Q: Why does AI default to a mixed look?
A: Training data conflates Southeast Asian, Asian-American, and mixed faces under one “Asian” label. Naming the country plus the eye structure (monolid / double eyelid / epicanthic fold) anchors the model precisely.
Q: Can I tell Chinese / Japanese / Korean output apart?
A: Partially. Korean leans K-pop, Japanese leans natural and editorial, and Chinese carries regional variance you may need to specify (Southern / Northern, etc.).
Q: How do I prompt mature East Asian women?
A: Name the age (35-year-old, 45-year-old) and add mature, calm, light wrinkles around eyes. Do not avoid age words; the model needs them.
Q: My skin keeps coming out too pale. Why?
A: fair skin alone biases cool-white. Add warm undertones, natural light tan, or light olive skin to correct it.
Q: How do I keep the same face across scenes?
A: In Midjourney v7, use the Omni Reference tab with a single clean front-facing photo at strength 300-500 (the older --cref does not run in v7). On Stable Diffusion XL, use InstantID or IP-Adapter-FaceID, or ReActor for a straight face swap. The prompt controls the environment, not the identity. Full workflow in How to Create Consistent AI Character Images.
Related articles
- Realistic Beautiful Woman Portrait Prompts
- Natural-Light Beautiful Woman Portrait Prompts
- Luxury Editorial Woman Portrait Prompts
- Chinese Aesthetic Character AI Prompts
- Midjourney Beginner Guide
- All AI Image Prompts
External reference: Midjourney’s Omni Reference documentation covers strength settings and reference-image requirements.