Lesson 13: Inpainting & Outpainting for Flawless Edits & Extensions

Lesson 13 covers inpainting and outpainting — the targeted editing techniques that fix imperfections and expand compositions after initial generation. These methods allow surgical corrections to hands, faces, explicit areas, or background elements while preserving the rest of the image, achieving seamless professional results impossible with base generation alone.

Why Inpainting/Outpainting Is Critical for Elite NSFW

• Fix common artifacts: deformed hands, fused fingers, asymmetrical faces, bad genitalia, extra limbs
• Refine explicit details: adjust wetness, texture, proportions without regenerating entire image
• Extend scenes: add more background, change environment, or create wider compositions
• Maintain consistency: preserve lighting, skin tone, and style across edits
• High control: mask only problem areas → denoise strength 0.5–0.8 for natural blending

Key Inpainting Settings & Best Practices

Inpainting Workflow in ComfyUI

1. Start from your saved pro txt2img workflow (Lesson 10).
2. Generate base image → right-click output → "Save Image" or keep in workflow.
3. Add Load Image node → load the base generation.
4. Add Mask Editor or Load Mask: Draw precise mask over problem area (hands, face, explicit zone) using ComfyUI's built-in mask tool or external (Photoshop/GIMP) and load PNG mask (black=keep, white=replace).
5. Add VAE Encode (for Inpaint) node → connect image and mask.
6. Add Inpaint Model Conditioning or use KSampler with mask input (connect mask to KSampler's mask input if using inpaint mode).
7. Use same checkpoint (HiDream-I1 uncensored).
8. Prompt: Keep original or add targeted fixes e.g., "perfect hands, detailed fingers, natural pose" (focus prompt on masked area).
9. Negative: Same Bible + emphasize fix e.g., "deformed hands, fused fingers".
10. KSampler settings: Denoise 0.6–0.8, Steps 40–60, CFG 6.0.
11. VAE Decode → Save Image (or preview).

Outpainting Workflow (Extending Canvas)

1. Use same base image loaded.
2. Add Image Pad for Outpaint node or manual crop/expand canvas with transparent borders.
3. Create mask covering the new/empty areas (white on extended parts).
4. Follow inpainting steps but use lower denoise (0.4–0.65) and prompt focused on new content (e.g., "extend luxurious bedroom background, large windows, city view at dusk").
5. Run multiple passes if needed for seamless extension.

Practical Fixes for Common NSFW Issues

• Bad hands: Mask hands only → denoise 0.75 → prompt "perfect detailed hands, five fingers each, natural pose".
• Face asymmetry: Mask face → denoise 0.6 → prompt "symmetrical beautiful face, detailed eyes".
• Explicit area deformities: Precise small mask → denoise 0.65–0.8 → targeted prompt "detailed realistic genitalia, natural folds, realistic wetness".
• Composition too tight: Outpaint sides/bottom → denoise 0.5 → prompt describing extended scene.

Best Practices & Troubleshooting

• Always mask slightly larger than problem area for better blending.
• Use same seed + low denoise for minimal changes; higher denoise for redraws.
• Iterate: Inpaint once → review → inpaint again if needed.
• Artifacts at edges: Increase mask blur/padding or run second low-denoise pass.
• Save inpaint-specific workflows (e.g., "Hand_Fix_Inpaint.json").

Assignment

1. Take 3–4 of your best images from Lesson 10 or 11 (with minor flaws like hands, face, or tight crop).
2. Build an inpainting workflow: Load image → mask problem area → inpaint with targeted prompt/negative.
3. Fix at least one issue per image (e.g., hands on one, explicit detail on another, face on third).
4. Try one outpainting extension on a strong vertical composition.
5. Save before/after pairs and final fixed versions.
6. Evaluate: Seamless blending? Improved realism? No visible seams or style shifts?

These inpainted/outpainted images are now near-final quality. Next lessons add upscaling and detail enhancement to reach true 4K+ professional output.

End of Lesson 13