Update README.md
Browse files
README.md
CHANGED
|
@@ -15,3 +15,52 @@ tags:
|
|
| 15 |
- blender
|
| 16 |
- 3d
|
| 17 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
- blender
|
| 16 |
- 3d
|
| 17 |
---
|
| 18 |
+
|
| 19 |
+
💎 DiamondForge XL (SDXL 1.0)
|
| 20 |
+
|
| 21 |
+
DiamondForge XL is an anime–realism hybrid SDXL checkpoint designed for clean character rendering, expressive faces, and cinematic lighting while maintaining strong prompt adherence across a wide range of scenarios.
|
| 22 |
+
|
| 23 |
+
Built and stress-tested through extensive prompt validation, this model balances stylization and realism, making it versatile for portraits, fashion, fantasy, and environmental compositions.
|
| 24 |
+
|
| 25 |
+
✨ Key Strengths
|
| 26 |
+
Stable facial structure across angles and expressions
|
| 27 |
+
Clean linework + soft shading balance (no overbaked textures)
|
| 28 |
+
Strong hair rendering (flow, highlights, and color fidelity)
|
| 29 |
+
Consistent lighting response from soft natural light to harsh directional setups
|
| 30 |
+
Good depth handling in both simple and complex environments
|
| 31 |
+
Resistant to style hijacking (holds prompt intent well)
|
| 32 |
+
🧪 Stress-Test Coverage
|
| 33 |
+
|
| 34 |
+
This model has been validated against:
|
| 35 |
+
|
| 36 |
+
Clothing edge cases (loose / layered garments)
|
| 37 |
+
Hair complexity (long, flowing, multi-tone)
|
| 38 |
+
Harsh and directional lighting scenarios
|
| 39 |
+
Style drift / unintended painterly bias
|
| 40 |
+
Environmental depth and perspective
|
| 41 |
+
Expression variation and facial consistency
|
| 42 |
+
Accessory stacking and detail density
|
| 43 |
+
Non-female subject sanity checks
|
| 44 |
+
⚙️ Recommended Settings
|
| 45 |
+
Resolution: 832×1216 (optimal balance of detail and stability)
|
| 46 |
+
Sampler: DPM++ 2M Karras / similar
|
| 47 |
+
Steps: 25–35
|
| 48 |
+
CFG: 4.5 – 7 (sweet spot around ~5.5)
|
| 49 |
+
Clip Skip: 1–2 depending on style preference
|
| 50 |
+
🎯 Use Cases
|
| 51 |
+
Character portraits
|
| 52 |
+
Stylized realism / anime hybrids
|
| 53 |
+
Fashion/editorial compositions
|
| 54 |
+
Fantasy / sci-fi characters
|
| 55 |
+
Animation pipelines (SDXL → WAN workflows)
|
| 56 |
+
🎥 Workflow Compatibility
|
| 57 |
+
|
| 58 |
+
Optimized for downstream use in:
|
| 59 |
+
|
| 60 |
+
Image-to-Video pipelines (WAN / AnimateDiff-style workflows)
|
| 61 |
+
Multi-phase generation setups
|
| 62 |
+
High-consistency character sequences
|
| 63 |
+
⚠️ Notes
|
| 64 |
+
Tuned primarily for female character generation, but supports broader subjects
|
| 65 |
+
Designed to avoid over-texturing artifacts seen in multi-pass workflows
|
| 66 |
+
Works best with clear prompt structure and moderate CFG values
|