Photoreal Portrait: cyberpunk samurai with glowing katana with colored gel split magenta and teal (Midjourney niji v6)

You are generating a photorealistic portrait prompt tuned for Midjourney niji v6. Deliver studio-quality output.

## Subject Sheet
- **Who/what:** cyberpunk samurai with glowing katana
- **Style anchor:** Pixar Disney 3D
- **Light:** colored gel split magenta and teal
- **Composition:** triangular composition
- **Aspect:** --ar 21:9

## Photoreal Portrait Formula
Real-photograph prompts follow this stack. Skip any layer and the output collapses toward illustration:

1. **Camera body + lens + aperture**
   - Examples: "Hasselblad X2D 100C, 90mm f/2.5 lens, f/2.0", "Canon R5, RF 85mm f/1.2L, wide open", "Leica M11, Summilux 50mm, f/1.4"
2. **Film stock or sensor aesthetic**
   - Digital clean, Kodak Portra 400, Cinestill 800T, Fuji Pro 400H, Tri-X 400 B&W
3. **Lighting specification**
   - colored gel split magenta and teal — name the modifier: softbox, beauty dish, strip light, natural north-facing window
4. **Subject's physical details**
   - Age, skin texture (NOT "flawless" — say "visible skin pores, fine lines around eyes, subtle shine on forehead")
   - Wardrobe in 3-part spec: color + material + silhouette
5. **Micro-expression or gesture**
   - "Slight smirk, eyes catching light off-camera left" beats "smiling"
6. **Environment cue**
   - One clause on background — out-of-focus bokeh or defined context
7. **Color grade**
   - Teal and orange, warm desaturated, clean natural, muted pastel

## Midjourney niji v6-Specific Photoreal Tips

- **Flux.1 [pro]:** The current leader for portraits. Skin looks real without "plastic face". Do NOT use tag soup — write prose.
- **Midjourney v6.1 / v7:** Must include `--style raw` to avoid its default stylization. `--stylize 50-150` for editorial realism.
- **Imagen 3:** Best-in-class skin texture via Gemini API. Responds to real camera/lens specs.
- **SDXL / SD 3.5:** Needs trigger LoRAs (e.g. RealVisXL, Juggernaut XL) to avoid the SDXL "plastic face". CFG 4-6.
- **DALL-E 3:** Sometimes adds cartoon exaggeration — counter with "natural photograph, unretouched, photojournalistic".
- **Playground v3:** Exceptional at editorial fashion portraits.
- **Nano Banana:** Fast iteration; great for testing composition before final render on Midjourney niji v6.

## Output

**A. PORTRAIT PROMPT** (70-130 words):
```
[Full prompt including all 7 layers]
```

**B. PARAMETERS:**
- Aspect: --ar 21:9
- Midjourney: `--style raw --stylize 100 --ar 21:9`
- Flux: [dev] or [pro], guidance 3.5, 25-30 steps
- SDXL: CFG 4.5, DPM++ 2M Karras, 30 steps, RealVisXL checkpoint
- Negative (SDXL only): `plastic skin, airbrushed, over-saturated, cartoon, 3D render, waxy`

**C. VARIATIONS (3 alternates):**
- **Same subject, different lens** — give an 85mm / 35mm / macro variant
- **Same subject, different crop** — propose waist-up vs extreme close-up
- **Same subject, different mood** — one lighting swap

**D. COMMON FAILURE MODES:**
1. Glossy "plastic" skin → add "visible skin pores, natural oil sheen, 10% imperfection" and lower the stylization/CFG.
2. Wall-eye or cross-eye → state "eyes focused on [specific point]".
3. Jewelry/hands disasters → specify hands explicitly or crop them out.
4. Overly symmetrical face → add "slight asymmetry, natural human proportions".

## Strict Rules
- NEVER say "beautiful", "stunning", "masterpiece" — these are tells that produce AI-aesthetic slop on modern models.
- Do not stack more than 5 style adjectives.
- If cyberpunk samurai with glowing katana is a real public figure, the major APIs will refuse — pivot to "reminiscent of [era/archetype]".
- Keep the prompt focused on ONE subject. Multi-subject photoreal portraits fail on all models below ~v7.

Generate the complete output now.