Text-to-Image AI Tools Without Content Filters

# Cyberpunk Bounty Hunter Portrait — Prompt Package

## 1. Recommended Model / Tool

**Primary pick: Flux.1-dev** (via ComfyUI or Forge) — best for photorealistic skin, accurate lighting, and sharp cinematic detail with minimal prompt-wrangling.

**Strong alternatives:**
- **SDXL + Juggernaut XL v9 / RealVisXL V4.0** — excellent gritty realism, widely supported in Automatic1111/Forge/ComfyUI.
- **epiCRealism / CyberRealistic (SD 1.5)** — if you want grainier, filmic output on lower VRAM.
- **Pony Diffusion V6 XL** — only if you want a more stylized/illustrated take (needs score tags).

Run locally in **ComfyUI** (best control) or **Forge/Automatic1111** (easiest). Fooocus works if you prefer a one-click workflow.

---

## 2. Positive Prompt

**For Flux.1-dev (natural language works best):**
```
Cinematic photorealistic portrait of a tough female bounty hunter in her early 30s standing in a narrow neon-lit Tokyo back alley at night. She has sharp piercing eyes, a weathered face with a diagonal scar across her cheek, faint cybernetic implant along her temple, messy shoulder-length dark hair with undercut, subtle smudged eyeliner. She wears a battered black leather jacket, torn at the shoulder, exposing a shoulder holster and a worn tactical undershirt. Wet asphalt reflecting pink and cyan neon signs, Japanese kanji signage glowing in the background, steam rising from a manhole, light rain. Dramatic rim lighting from a magenta neon sign behind her, cool cyan fill light on her face, deep shadows, shallow depth of field, 85mm lens, shot on Arri Alexa, anamorphic bokeh, Blade Runner 2049 color grading, gritty cinematic atmosphere, ultra-detailed skin texture, photorealistic, 8k.
```

**For SDXL / SD1.5 (tag-weighted version):**
```
(masterpiece:1.2), (photorealistic:1.4), cinematic portrait, female bounty hunter, early 30s, tough expression, (scarred face:1.2), diagonal cheek scar, piercing eyes, subtle cybernetic temple implant, messy dark hair with undercut, (torn black leather jacket:1.3), shoulder holster, tactical undershirt, fingerless gloves, 
neon-lit Tokyo back alley, night, wet asphalt, reflections, (magenta and cyan neon signs:1.2), Japanese kanji signage, steam, light rain, 
(dramatic rim lighting:1.3), backlit neon, cool cyan fill, deep shadows, volumetric haze, shallow depth of field, bokeh, 85mm, anamorphic, 
Blade Runner aesthetic, cyberpunk, gritty, moody, film grain, (hyperdetailed skin:1.2), pore-level detail, sharp focus, 8k, Kodak Portra, cinematic color grade
```

---

## 3. Negative Prompt

**SDXL / SD1.5:**
```
(worst quality:1.4), (low quality:1.4), (normal quality:1.2), lowres, blurry, out of focus, jpeg artifacts, compression artifacts, 
cartoon, anime, 3d render, cgi, painting, illustration, sketch, plastic skin, doll-like, airbrushed, overexposed, flat lighting,
deformed, disfigured, bad anatomy, extra limbs, extra fingers, missing fingers, mutated hands, fused fingers, long neck, asymmetrical eyes, crossed eyes, malformed face,
text, watermark, signature, logo, frame, border, cropped, 
clean clothes, pristine, cheerful, bright daylight, sunny, washed out colors
```

*(Flux.1-dev doesn't use a negative prompt — just omit undesired elements from the positive prompt, or use a negative via ComfyUI nodes with low weight.)*

---

## 4. Suggested Settings

### Flux.1-dev
| Setting | Value |
|---|---|
| Sampler | Euler / DPM++ 2M |
| Scheduler | Simple / Beta |
| Steps | 25–30 |
| Guidance | 3.5 |
| Resolution | 896×1152 or 1024×1280 (portrait) |
| Tip | Use the FP8 or GGUF quant if VRAM-limited |

### SDXL (Juggernaut XL / RealVisXL)
| Setting | Value |
|---|---|
| Sampler | DPM++ 2M Karras (or DPM++ 3M SDE Karras) |
| Steps | 30–40 |
| CFG | 4.5–6 |
| Resolution | 832×1216 (portrait) or 896×1152 |
| Refiner | Optional, 0.2 denoise |
| Hires fix | 1.5× upscale, 4x-UltraSharp, denoise 0.35 |

### SD 1.5 (epiCRealism / CyberRealistic)
| Setting | Value |
|---|---|
| Sampler | DPM++ 2M Karras |
| Steps | 28–35 |
| CFG | 6–7 |
| Resolution | 512×768, then Hires fix 2× (R-ESRGAN 4x+ or 4x-UltraSharp, denoise 0.4) |

---

### Extra tips
- **Add LoRAs** like *Cyberpunk Edgerunners style*, *Detail Tweaker XL*, or *Cinematic Lighting* at 0.3–0.6 weight for more flavor.
- **ADetailer / FaceDetailer** pass on the face drastically improves eyes, scars, and skin texture.
- Run 4–8 seeds, pick the strongest composition

# Cyberpunk Bounty Hunter Portrait Prompt Package

## Recommended Model & Tools
**Primary:** SDXL 1.0 or **Juggernaut XL** (excellent for gritty realism + character detail)
**Alternative:** Flux.1-dev (superior lighting/composition but slower)
**Hosting:** Automatic1111 (local), ComfyUI (advanced control), Fooocus (user-friendly), or Mage.space (free hosted)

---

## Positive Prompt

```
a tough, scarred female bounty hunter standing in a neon-lit Tokyo alley at night, 
cyberpunk aesthetic, dramatic rim lighting with cyan and magenta neon reflections, 
wearing a distressed black leather jacket with tears and battle damage, 
weathered face with prominent scars, intense expression, gritty realism, 
cinematic noir mood, rain-slicked pavement, holographic signs in background, 
shallow depth of field, volumetric god rays cutting through mist, 
shot on Kodak film, noir color grading, professional portrait photography, 
8k resolution, hyper-detailed, photorealistic, dramatic side lighting, 
moody and atmospheric, tension-filled composition
```

---

## Negative Prompt

```
soft lighting, bright cheerful, cartoon, anime, illustration style, 
blurry, low quality, distorted face, bad anatomy, asymmetrical features, 
plastic skin, oversaturated colors, lens flare artifacts, motion blur, 
watermark, text, low resolution, poorly lit, flat lighting, 
soft focus background, generic features, too young appearance, 
clean/pristine condition, artificial, CGI plastic look
```

---

## Recommended Settings (SDXL)

| Parameter | Value |
|-----------|-------|
| **Sampler** | DPM++ 2M Karras or Euler |
| **Steps** | 30–40 (SDXL is efficient) |
| **CFG Scale** | 7.5–8.5 (strong adherence without over-saturation) |
| **Resolution** | 1024×1024 or 896×1152 (portrait crop) |
| **Seed** | Random or lock for iteration |
| **Upscaler** | RealESRGAN 4x (post-process) |

---

## Alternative Positive Prompt (Flux-optimized)

```
cinematic photograph: scarred female bounty hunter in Tokyo cyberpunk alley, 
neon cyan and magenta rim lighting, torn leather jacket, hardened expression, 
rain-wet streets, holographic advertisements, volumetric atmosphere, 
professional film noir color grading, dramatic side-key lighting, 
photorealistic skin texture detail, moody dystopian mood, 
shot on Fujifilm Velvia, 8k detail, shallow DOF
```

---

## Pro Tips

- **Iteration:** Lock a good seed and vary CFG (7–9) to refine mood
- **Post-process:** Apply slight desaturation + crush blacks in Photoshop for extra noir feel
- **Lighting refinement:** If rim lighting is weak, add "strong rim light" and increase CFG slightly
- **Detail enhancement:** Use an upscaler (RealESRGAN, Topaz) and inpaint face details if needed
- **Composition:** Specify "rule of thirds" or "off-center framing" if generation feels centered

This setup targets gritty realism with strong cinematic mood—excellent for character-driven cyberpunk work.

## Recommended model / tool
**Best overall:** **SDXL** in **Automatic1111**, **ComfyUI**, or **Fooocus**  
Why: strong realism, cinematic lighting, detailed environments, and good prompt adherence for gritty cyberpunk portraits.

**Good alternatives**
- **Flux.1-dev**: excellent realism and prompt understanding
- **Juggernaut XL** or **RealVisXL**: strong photoreal/cinematic SDXL checkpoints
- **Stable Diffusion 1.5 + realistic cyberpunk checkpoint**: lighter hardware, but less coherent than SDXL

---

## Positive prompt
Use this as a base prompt:

**Prompt:**
> gritty realistic cyberpunk portrait of a female bounty hunter, standing in a narrow neon-lit Tokyo alley at night, tough expression, visible facial scars, weathered skin, intense eyes, short messy dark hair, torn black leather jacket, tactical layers, subtle futuristic implants, worn gloves, urban survivor aesthetic, rain-slicked alley, glowing Japanese signage, holographic advertisements, steam rising from vents, wet pavement reflections, cinematic composition, medium close-up, slightly low angle, dramatic rim lighting, neon magenta and cyan edge light, moody shadows, high contrast, volumetric mist, realistic skin texture, detailed fabric wear, photorealistic, ultra-detailed, grounded sci-fi, dystopian atmosphere, shallow depth of field, sharp focus on face, filmic color grading, cinematic still, high detail, masterpiece

### Optional stronger cinematic version
> photorealistic cinematic portrait of a battle-hardened female bounty hunter in a neon-soaked Tokyo back alley at night, scarred face, stern gaze, rugged beauty, torn leather jacket with scratches and rain damage, layered tactical clothing, cyberpunk realism, subtle tech enhancements, dramatic magenta and teal rim light, backlit by neon signs, wet asphalt reflecting colored light, drifting steam, cluttered alley details, cables, vending machines, graffiti, moody atmospheric haze, realistic pores, natural facial asymmetry, 85mm lens look, shallow depth of field, high dynamic range, gritty urban noir, ultra detailed, cinematic lighting, professional color grading, realistic composition

---

## Negative prompt
> anime, cartoon, illustration, painting, lowres, blurry, soft focus, out of frame, bad anatomy, deformed face, asymmetrical eyes, extra fingers, missing fingers, fused hands, malformed hands, duplicate person, multiple heads, poorly drawn jacket, plastic skin, over-smoothed skin, doll face, oversaturated, flat lighting, washed out, low contrast, jpeg artifacts, text, watermark, logo, bad perspective, mutated limbs, unrealistic scars, excessive cybernetics, cluttered composition, noisy image, overexposed highlights, underexposed face

---

## Suggested settings

### For SDXL
- **Sampler:** DPM++ 2M Karras
- **Steps:** 30–40
- **CFG:** 5.5–7
- **Resolution:** **1024x1280** or **832x1216**
- **Hires fix:** optional, 1.5x upscale
- **Denoise:** 0.25–0.4

### For Flux.1-dev
- **Steps:** 20–30
- **Guidance/CFG:** lower than SDXL, around **3–4.5** depending on UI
- **Resolution:** **1024x1280**
- Use natural-language prompting; Flux often performs better with less tag stuffing

### For SD 1.5
- **Sampler:** DPM++ SDE Karras or Euler a
- **Steps:** 28–40
- **CFG:** 7–9
- **Resolution:** **768x1024**
- Consider **Hires fix** to improve facial detail

---

## Prompt tuning tips
If you want more realism, add:
> photojournalistic, realistic imperfections, natural skin pores, subtle makeup, believable scars, practical wardrobe

If you want more cyberpunk atmosphere, add:
> dense neon fog, holograms, rain, kanji signs, electric reflections, urban decay, tech-noir

If you want a stronger portrait focus, add:
> centered subject, face emphasis, shoulder-up portrait, 85mm lens, shallow DOF

If you want her to feel more dangerous, add:
> hardened veteran, cold stare, battle scars, confident posture, predator-like calm

---

## Compact paste-ready package

**Model:** SDXL / Juggernaut XL / RealVisXL  
**Prompt:**
> gritty realistic cyberpunk portrait of a female bounty hunter in a neon-lit Tokyo alley at night, tough scarred face, intense eyes, messy dark hair, torn black leather jacket, layered tactical clothing, subtle futuristic implants, rain-soaked alley, glowing Japanese neon signs, holograms, steam, wet pavement reflections, medium close-up, slightly low angle, dramatic magenta and cyan rim lighting, moody shadows, cinematic atmosphere, high contrast, volumetric mist, realistic skin texture, detailed clothing damage, shallow depth of field, sharp focus on face, filmic color grading, photorealistic, ultra-detailed, cinematic still, masterpiece

**Negative:**
> anime, cartoon, painting, lowres, blurry, bad anatomy, deformed face, asymmetrical eyes, extra fingers, malformed hands, duplicate person, plastic skin, oversmoothed skin, oversaturated, flat lighting, watermark, logo, text, jpeg artifacts, unrealistic scars, excessive cybernetics, noisy image

**Settings:**
> DPM++ 2M Karras, 35 steps, CFG 6.5, 1024x1280

If you want, I can also generate:
1. a **more photoreal version**,  
2. a **darker Blade Runner-style version**, or  
3. a **ComfyUI node-ready prompt workflow**.

Of course. Here is a complete prompt package optimized for unrestricted, open-source models

### Recommended Model/Tool
**SDXL 1.0** (via Automatic1111 WebUI or ComfyUI local install) with a realistic checkpoint like Juggernaut XL or Realistic Vision XL fine-tune. For superior text adherence and photorealism, use **Flux.1-dev** (via ComfyUI or hosted on fal.ai/Replicate). Both are uncensored in local setups. Avoid filtered hosted services.

### Positive Prompt
```
gritty photorealistic cyberpunk portrait of a tough female bounty hunter, mid-30s, intense piercing eyes, detailed facial scars and cybernetic implants, weathered skin, short messy black hair, smirking expression, wearing torn black leather jacket with patches and spikes, high collar, subtle tattoos on neck, neon-lit narrow Tokyo alley at night, rain-slicked wet pavement reflecting vibrant pink blue purple neon signs, holographic ads and flickering billboards in background, dense fog and steam rising, dramatic rim lighting from behind with god rays piercing through, cinematic composition, shallow depth of field, volumetric god rays, high contrast, ultra-detailed skin texture, sharp focus, 8k, masterpiece, award-winning photography, cyberpunk 2077 style, blade runner atmosphere
```

### Negative Prompt
```
blurry, lowres, low quality, deformed, mutated, ugly, disfigured, poorly drawn face, poorly drawn hands, extra limbs, missing limbs, extra fingers, fused fingers, malformed limbs, bad anatomy, watermark, text, signature, logo, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, compression artifacts, out of frame, mutated hands, mutation, ugly face, extra arms, extra legs
```

### Suggested Settings
- **Sampler**: DPM++ 2M Karras (SDXL) or Euler (Flux)
- **Steps**: 30-40 (SDXL) or 20-28 (Flux)
- **CFG Scale**: 7-8.5
- **Resolution**: 768x1152 (portrait orientation) or 1024x1024 (square, then crop)
- **Seed**: -1 (random) or experiment with fixed for variations
- **Other**: Hires fix enabled (upscale by 1.5x, denoise 0.3-0.5); VAE: sd_xl_vae.safetensors (for SDXL); Clip skip: 1-2

Paste directly into your generator. Refine weights like (scarred face:1.2) for emphasis if needed. Generate 4+ images per run for best picks.