Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: agpl-3.0
|
| 3 |
+
tags:
|
| 4 |
+
- anime
|
| 5 |
+
- image-generation
|
| 6 |
+
- sdxl
|
| 7 |
+
- stable-diffusion
|
| 8 |
+
- 3d
|
| 9 |
+
- deterministic
|
| 10 |
+
- dora
|
| 11 |
+
- text-to-image
|
| 12 |
+
- noobai
|
| 13 |
+
language:
|
| 14 |
+
- en
|
| 15 |
+
pipeline_tag: text-to-image
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# 4CGT — 四彫画匠 (Shichō Gashō)
|
| 19 |
+
|
| 20 |
+
**Master Artisan of Sculpted Art** — quality-engineered anime image generation with cross-platform determinism, DoRA anatomical accuracy, and integrated 2D→3D conversion.
|
| 21 |
+
|
| 22 |
+
**GitHub**: [teenu/4cgt](https://github.com/teenu/4cgt)
|
| 23 |
+
**Website**: [4cgt.com](https://4cgt.com)
|
| 24 |
+
|
| 25 |
+
## Assets
|
| 26 |
+
|
| 27 |
+
This repo contains all assets needed to run 4CGT:
|
| 28 |
+
|
| 29 |
+
| File | Size | Description |
|
| 30 |
+
|------|------|-------------|
|
| 31 |
+
| `NoobAI-XL-Vpred-v1.0.safetensors` | 6.7 GB | NoobAI XL V-Pred 1.0 model (BF16) |
|
| 32 |
+
| `controlnet/openpose_fp32.safetensors` | 4.7 GB | OpenPose ControlNet for pose control |
|
| 33 |
+
| `dora/noobai_vp10_stabilizer_v0.280a_fp16.safetensors` | 64 MB | DoRA stabilizer v0.280a (recommended) |
|
| 34 |
+
| `dora/noobai_vp10_stabilizer_v0.271_fp16.safetensors` | 44 MB | DoRA stabilizer v0.271 |
|
| 35 |
+
| `style/*.csv` | 109 MB | Danbooru + e621 character/artist databases |
|
| 36 |
+
| `poses/*.png` | 140 KB | Example OpenPose skeleton images |
|
| 37 |
+
|
| 38 |
+
## Quick Start
|
| 39 |
+
|
| 40 |
+
```bash
|
| 41 |
+
# 1. Clone the tool
|
| 42 |
+
git clone https://github.com/teenu/4cgt.git
|
| 43 |
+
cd 4cgt
|
| 44 |
+
|
| 45 |
+
# 2. Install PyTorch (pick your GPU)
|
| 46 |
+
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124
|
| 47 |
+
|
| 48 |
+
# 3. Install dependencies
|
| 49 |
+
pip install -r requirements.txt
|
| 50 |
+
|
| 51 |
+
# 4. Download assets from this repo
|
| 52 |
+
pip install huggingface-hub
|
| 53 |
+
huggingface-cli download epigene/4cgt --local-dir assets
|
| 54 |
+
|
| 55 |
+
# 5. Move assets into place
|
| 56 |
+
mv assets/NoobAI-XL-Vpred-v1.0.safetensors .
|
| 57 |
+
mv assets/dora/* dora/
|
| 58 |
+
mv assets/controlnet/* controlnet/
|
| 59 |
+
mv assets/style/* style/
|
| 60 |
+
mv assets/poses/* poses/
|
| 61 |
+
|
| 62 |
+
# 6. Run
|
| 63 |
+
python main.py
|
| 64 |
+
```
|
| 65 |
+
|
| 66 |
+
## Four Pillars
|
| 67 |
+
|
| 68 |
+
- **Anatomical Accuracy** — DoRA weight-decomposed adapters optimized for v-prediction
|
| 69 |
+
- **Prompt Fidelity** — Long prompt support up to 600 tokens with A1111-compatible syntax
|
| 70 |
+
- **Image Quality** — Lossless FP32 VAE, BF16/FP32 only, FP16 rejected by design
|
| 71 |
+
- **Deterministic Output** — Same seed = identical image hash across macOS, Windows, Linux
|
| 72 |
+
|
| 73 |
+
## License
|
| 74 |
+
|
| 75 |
+
AGPL-3.0. See [GitHub](https://github.com/teenu/4cgt) for full license.
|