Face Restoration and Enhancement

Restore and enhance faces in SD generations.

**Face Issue:** ${{ISSUE}}
**Enhancement Goal:** ${{GOAL}}
**Quality Level:** ${{QUALITY}}
**Style Preservation:** ${{PRESERVATION}}

Face restoration methods:

**Built-in Options:**

CodeFormer:
- Adjustable fidelity
- Preserves features
- Good for mild issues
- Settings: weight 0.5-0.8

GFPGAN:
- Strong restoration
- May alter features slightly
- Good for severe issues
- Settings: visibility 0.5-1.0

**When to Use What:**

CodeFormer better for:
- Preserving likeness
- Subtle enhancement
- Artistic styles
- High fidelity needs

GFPGAN better for:
- Severe distortion
- Quick fixes
- General improvement
- Realistic styles

**ADetailer Extension:**
Automated face fixing:
- Detects faces automatically
- Runs inpainting on faces
- Customizable prompts
- Multiple detection models

ADetailer Settings:
- Model: face_yolov8n.pt
- Confidence: 0.3
- Mask padding: 32
- Denoising: 0.4

**Manual Face Fix:**

Inpainting workflow:
1. Mask face area
2. Use face-focused prompt
3. Low denoising (0.3-0.5)
4. Match style

Prompt for face inpaint:
detailed face, perfect face,
symmetrical features,
(same style as image)

**Prevention Tips:**
- Use face-focused prompts
- Include face quality tags
- Use face-optimized checkpoints
- Generate at higher resolution

**Common Issues:**
- Asymmetric: Use face restore
- Distorted: Inpaint
- Wrong features: Regenerate
- Blurry: Upscale with face fix