Skip to content
All guides
Technical MasterclassInteractivePortrait PhotographyTechnical

Asian Portraits: Nationality Differences

Explore how AI models handle the visual distinctions between Asian nationalities. This guide examines what models can capture accurately versus where they default to generic features—because "Asian" encompasses billions of people across vastly different cultures and physical characteristics.

A Technical Explorationmodel=nano-banana-pro

Why "Asian" Is Too Broad for AI

A note before we begin: This guide explores physical characteristics that AI models may or may not capture accurately. These are generalizations—individuals within any nationality vary enormously. The goal isn't to stereotype but to understand AI's technical capabilities and limitations when generating specific ethnicities.

When you prompt an AI model with "Asian woman," you're asking it to represent over 4.7 billion people across 48 countries with vastly different physical characteristics. The result is typically a generic, averaged face that represents no one authentically—often defaulting to a narrow subset of East Asian features.

Professional use cases demand specificity: casting directors need distinct nationalities, stock photographers need authentic representation, character designers need visual accuracy. This guide demonstrates what's achievable with precise prompting—and where current models still fall short.

We're using Nano Banana Pro for this exploration. Judge the results yourself—where does the model succeed in capturing nationality-specific features, and where does it fall back on generic Asian defaults?

Foundation

Eight Nationalities, One Word

When AI hears 'Asian,' it produces an averaged face. Here's what specificity achieves—eight distinct nationalities, each with their own visual identity.

JapanesePortrait of a Japanese woman in her 20s, cool pink undertones, soft roun…
KoreanPortrait of a Korean woman in her 20s, high cheekbones, V-shaped jaw, fa…
ChinesePortrait of a Han Chinese woman in her 20s, warm yellow undertones, broa…
VietnamesePortrait of a Vietnamese woman in her 20s, warm tan golden skin, oval fa…
ThaiPortrait of a Thai woman in her 20s, golden-brown skin, rounder face, pr…
FilipinoPortrait of a Filipina woman in her 20s, warm brown skin, Malay features…
IndonesianPortrait of a Javanese woman in her 20s, warm medium-brown skin, delicat…
IndianPortrait of a Tamil woman in her 20s, deep bronze skin, large expressive…

Compare these eight portraits. Each represents a different nationality with distinct features: Japanese porcelain skin with cool undertones versus Thai golden-brown warmth. Korean angular structure versus Indonesian soft Malay features. The differences are significant—yet a generic "Asian woman" prompt would collapse all this diversity into a single averaged face.

The key insight: Nationality-specific prompting isn't about stereotyping—it's about honoring the actual diversity that exists. Generic prompts erase this diversity; specific prompts preserve it.

Create accurate diverse portraits

Access Nano Banana Pro and 40+ other models. Generate nationality-specific portraits with authentic features using a simple img tag.

Start Free Trial
East Asia

Japanese vs Korean vs Chinese

The most commonly conflated nationalities. Same lighting, same prompt structure—different nationality specifications reveal distinct features.

Japanese Woman

Japan

Korean Woman

South Korea

Northern Chinese Woman

Northern China

Southern Chinese Woman

Southern China

Japanese features often include softer, rounder contours, porcelain skin with cool pink undertones, and a delicate chin. Double eyelids are common.

Korean features are characterized by high prominent cheekbones, the famous V-shaped jawline, fair skin with neutral undertones, and a high nose bridge.

Chinese features vary significantly by region. Northern Chinese typically have fairer skin with yellow undertones and broader facial structures, while Southern Chinese often have warmer tan skin and softer features with Southeast Asian influence.

How well did the model capture these distinctions? The Japanese portrait should feel softer than the Korean; the Northern Chinese broader than the Japanese. These are subtle differences that require careful prompting.

Southeast Asia

Vietnamese, Thai, Filipino, Indonesian

Often underrepresented in AI training data. These four nationalities have distinct Malay-Austronesian heritage with regional variations.

Vietnamese Woman

Vietnam

Thai Woman

Thailand

Filipino Woman

Philippines

Indonesian Woman

Indonesia

Southeast Asian nationalities share Malay-Austronesian heritage but show distinct regional characteristics. Vietnamese features bridge East and Southeast Asian characteristics—warmer than Chinese but with narrower features than Thai or Filipino.

Thai features often include rounder faces, prominent cheekbones, golden-brown skin, and a distinctive eye shape with gentle upward tilt.

Filipino features blend Malay heritage with Spanish colonial influence, creating unique combinations—rounder faces and fuller lips with occasionally more angular noses.

Indonesian features, particularly Javanese, often show delicate Malay-Polynesian characteristics with graceful refined bone structure.

NoteAI models often struggle with Southeast Asian nationalities due to training data imbalances. If you're getting generic East Asian features, try adding regional context: "warm Southeast Asian climate," "tropical light," or reference specific regional photography styles.
South Asia

North Indian vs South Indian vs Bengali

The Indian subcontinent contains more diversity than all of Europe. Regional distinctions are dramatic and important.

Northern India

North Indian Woman

Southern India

South Indian Woman

West Bengal

Bengali Woman

North Indian features (Punjabi, Kashmiri) often show Indo-Aryan heritage with fair to light brown skin, sharp defined noses, and refined oval faces showing Persian-Central Asian influence.

South Indian features (Tamil, Telugu, Malayalam) reflect Dravidian heritage with deeper brown skin, broader noses, larger eyes, and fuller lips. Skin tones are significantly darker than North Indian.

Bengali features bridge North and East, with distinctive oval faces, the famous "fish-shaped" expressive eyes, and medium brown skin with warm olive undertones.

TipFor Indian portraits, always specify region: Punjabi, Tamil, Bengali, Gujarati. "Indian woman" is nearly as broad as "Asian woman"—India alone has 22 official languages and corresponding ethnic diversity.
Overview

East vs Southeast vs South Asian

A high-level comparison of the three major Asian regions, each with dramatically different characteristics.

East Asian

  • →Fairer skin tones (pink to yellow undertones)
  • →Single or double eyelid variations
  • →Higher cheekbones, narrower faces
  • →Smaller, narrower noses
  • →Thinner lips generally
  • →Straight black hair

Southeast Asian

  • →Warmer, tan to golden-brown skin
  • →Rounder eye shapes often with natural lid crease
  • →Wider, flatter noses
  • →Fuller lips
  • →Rounder face shapes
  • →Wavy or straight black hair

South Asian

  • →Wide range from fair to deep brown skin
  • →Large expressive eyes with thick lashes
  • →Defined noses varying by region
  • →Fuller lips especially in South India
  • →Diverse facial structures by ethnicity
  • →Wavy to curly thick black hair
Technique

Nationality-Specific Prompting

The vocabulary that works for each nationality. Reference these descriptors to get authentic rather than generic results.

Japanese

porcelain skin with cool pink undertones, soft rounded features, gentle high cheekbones

Korean

high prominent cheekbones, V-shaped jawline, fair skin with neutral undertones

Chinese (Northern)

fair skin with yellow undertones, broader flatter facial structure, single eyelid distinctive shape

Vietnamese

warm tan skin with golden undertones, oval face with soft contours, Vietnamese eye shape

Quick Reference: Key Descriptors

Japanese

Porcelain, cool pink, soft rounded, double eyelid

Korean

High cheekbones, V-shaped jaw, neutral fair, monolid

Vietnamese

Warm tan, golden undertones, oval face, full lips

Thai

Golden-brown, rounder face, upward eye tilt, prominent cheekbones

Filipino

Warm brown, Malay-Spanish blend, full lips, wavy hair

North Indian

Fair to light brown, sharp nose, Indo-Aryan features

South Indian

Deep bronze, large eyes, broader features, Dravidian

Indonesian

Medium brown, Malay-Polynesian, delicate features

Avoid

Common Pitfalls and Solutions

Where AI goes wrong with Asian nationalities—and how to fix it.

Generic Asian

Using 'Asian woman' produces averaged, non-specific features

Wrong Skin Tone

Models default to East Asian skin for all Asian nationalities

Feature Mixing

AI combines features from different Asian nationalities

Beauty Standard Bias

Models gravitate toward Korean beauty standards for all Asians

Generic Asian:

Using 'Asian woman' produces averaged, non-specific features. Solution: Specify nationality, regional heritage, and distinctive features.

Wrong Skin Tone:

Models default to East Asian skin for all Asian nationalities. Solution: Specify undertones explicitly for each nationality.

Feature Mixing:

AI combines features from different Asian nationalities. Solution: Be specific about bone structure, eye shape, and facial proportions.

Beauty Standard Bias:

Models gravitate toward Korean beauty standards for all Asians. Solution: Reference specific regional aesthetics and photography styles.

Mastery

Putting It All Together

A masterwork portrait demonstrating precise nationality-specific prompting with all the techniques from this guide.

Mixed Heritage: Thai-Chinese

"Masterwork portrait of a Thai-Chinese woman from Chiang Mai... warm golden-tan skin with distinctive yellow undertones... high cheekbones from Chinese ancestry combined with softer Thai contours... the unique Thai-Chinese blend common in Northern Thailand..."

Why this works

  • →Specific mixed heritage (Thai-Chinese)
  • →Regional context (Chiang Mai, Northern Thailand)
  • →Undertones explicitly described
  • →Blended features acknowledged and described
  • →Cultural context (Lanna, Thai silk)

Prompt structure

  • 1.Heritage + Region + Age
  • 2.Skin tone + Undertones + Heritage explanation
  • 3.Blended features described specifically
  • 4.Cultural elements (clothing, jewelry)
  • 5.Lighting + Technical specs
Practice

Create Your Nationality-Specific Portrait

Apply what you've learned. Specify nationality, undertones, distinctive features, and regional context for authentic results.

Portrait of a Vietnamese woman in her mid-20s, warm tan skin wit…

Frequently asked

Why do AI models struggle with Asian nationalities?Training data imbalances mean some nationalities are overrepresented while others are underrepresented. Models also tend toward 'averaged' features when given vague prompts. East Asian features, particularly Korean beauty standards, often dominate because of their prevalence in training data.
Is it offensive to describe physical differences between nationalities?This is a technical guide about AI capabilities, not a statement about how people should look. Physical variation exists within all populations, and individuals within any nationality vary enormously. The goal is accurate representation in AI-generated imagery, not stereotyping.
How accurate are these nationality-specific generations?Results vary. Some distinctions (like skin undertone) are captured well with explicit prompting. Others (like subtle bone structure differences) are harder for models to consistently render. The guide shows what's currently achievable—judge the accuracy for yourself.
What about mixed-heritage Asian individuals?Mixed heritage is common throughout Asia. You can describe specific combinations: 'Japanese-Filipino heritage with features blending East and Southeast Asian characteristics.' Results vary based on how well the model understands these combinations.
Why focus on physical features at all?For professional applications—casting, stock photography, character design, medical visualization—accuracy matters. A 'Japanese' character who looks generically 'Asian' fails to represent. Technical accuracy enables authentic representation.

Key takeaways

01Specify nationality, not continent "Asian" is too broad. Japanese, Korean, Vietnamese, Thai, and Indian all require different descriptors for authentic results.
02Undertones vary dramatically Japanese cool pink, Korean neutral, Thai golden-brown, Tamil deep bronze—skin undertones differ significantly across Asian nationalities.
03Bone structure tells the story High Korean cheekbones, softer Japanese contours, rounder Thai faces, Bengali oval shapes—facial structure is the key differentiator.
04Eye shape is nuanced Beyond 'Asian eyes,' there's Japanese double eyelid, Korean monolid, Vietnamese single lid, Thai upward tilt—each distinct.
05Reference regional photographers Rinko Kawauchi for Japanese, Bangkok editorial for Thai, Raghu Rai for Indian—regional aesthetics improve authenticity.
06Models have biases AI often defaults to East Asian features for all Asians and Korean beauty standards as the ideal. Explicit prompting counters this.

Diversity deserves accuracy.
Capture it faithfully.

Free 7-day trial included. Cancel any time.