DREAM TEXTURES - FULL WIKI DOCUMENTATION
Overview Repo: https://github.com/carson katri/dream textures Wiki: https://github.com/carson katri/dream textures/wiki Version: Latest Blender: 3.3+ License: GPL 3.0
📚 Референсы 2 блоков кода обновлено 2026-10-02
Overview
Repo: https://github.com/carson-katri/dream-textures Wiki: https://github.com/carson-katri/dream-textures/wiki Version: Latest | Blender: 3.3+ | License: GPL-3.0
Setup Guide (from Wiki)
Installation
- Download from latest release
- Windows: Extract 7-Zip archive → get
.zip→ Install in Blender Preferences → Add-ons - Enable Dream Textures in add-ons
- DO NOT manually install dependencies - use prebuilt version only
Download Models
In add-on preferences, search and download:
stabilityai/stable-diffusion-2-1-base- Primary 512x512 modelstabilityai/stable-diffusion-2-inpainting- For inpainting/outpaintingstabilityai/stable-diffusion-2-depth- For texture projection, depth-to-imagestabilityai/stable-diffusion-x4-upscaler- For 4x AI upscalingstabilityai/stable-diffusion-2-1- Fine-tuned for 768x768
Private Models
- Generate Hugging Face token at https://huggingface.co/settings/tokens
- Paste in "Token" field in preferences
Import Checkpoints (.ckpt)
- Click "Import Checkpoint File" in preferences
- Select file → specify Model Config:
v1- Original CompVis models (SD 1.x)v2 (512, epsilon)- 512x512 models with prediction_type="epsilon"v2 (768, v_prediction)- 768x768 models with prediction_type="v_prediction"
DreamStudio (Cloud)
- Enter API key from https://beta.dreamstudio.ai/membership?tab=apiKeys
- Processes in cloud if local hardware insufficient
Image Generation (from Wiki)
Access
- Open Image Editor or Shader Editor
- Press
Nfor sidebar → Dream panel
Options
| Option | Description |
|---|---|
| Pipeline | Stable Diffusion (local) or DreamStudio (cloud) |
| Model | Choose installed model (some features require specific models) |
| Prompt | Text prompt. Presets available (e.g., "Texture" adds "texture" suffix) |
| Seamless | Circular convolution for perfect tiling. Can specify axes |
| Negative | What to avoid (e.g., "building" to remove buildings) |
| Size | Target dimensions (multiples of 64). Best at 512x512 |
| Source Image | File or Open Image for img2img |
Source Image Modes
- Color - Mix image with noise
- Color + Generated Depth - MiDaS depth + color (requires depth model)
- Color + Depth Map - Custom depth map image
- Depth - Treat image as depth map only
Advanced Parameters
| Parameter | Description |
|---|---|
| Random Seed | Auto-generate or manual seed |
| Steps | Sampler steps (more = better convergence) |
| CFG Scale | Prompt adherence strength |
| Scheduler | Different convergence speeds |
| Step Preview | "Fast" (latent preview) or "Accurate" (VAE decode each step) |
| Speed Optimizations | Half Precision recommended |
| Memory Optimizations | Attention Slicing (Auto) recommended |
Texture Projection (from Wiki)
Workflow
- Select objects in 3D Viewport
- Enter Edit Mode → select faces to target
- Local View (
/) to limit depth map to selection - Tip: Subdivide large faces if projection warps
Settings
- Choose depth model (e.g.,
stable-diffusion-2-depth) - Enter prompt
- Data source: "Depth" (default) or "Depth and Color" (uses viewport shading)
- Adjust image size to match viewport aspect ratio
Execution
- Click Project Dream Texture
- New material created with seed name
- UVs updated for projection angle
- Live preview in Viewport Shading mode
Inpaint and Outpaint (from Wiki)
Inpainting
- Mask area in Image Editor
- Requires
stable-diffusion-2-inpaintingmodel - Prompt: "seamless continuation of [texture]"
- Making Textures Seamless: Inpaint edges to blend
Outpainting
- Extend image beyond boundaries
- Choose origin (center, corners, edges)
- Generates new content matching style
Render Engine / Node System (from Wiki)
Nodes Added
- Dream Texture Node - Generate textures in shader graph
- ControlNet Node - Depth/Canny/Normal guidance
- Upscale Node - AI upscaling in pipeline
Render Pass Integration
- Use Cycles render as ControlNet input
- Re-style via img2img
- Animation: frame-by-frame re-styling
Controls
| Parameter | Effect |
|---|---|
| Depth | Use depth pass for structure |
| Noise Strength | Deviation from render (0-1) |
| Seed | Reproducibility |
AI Upscaling (from Wiki)
- 4x upscale via ESRGAN/RealESRGAN
- Tile Size option for large images
History (from Wiki)
- Recall, export, import generations
- Export: Save prompt + params + seed
- Import: Restore previous generation settings
Development Setup (from Wiki)
git clone https://github.com/carson-katri/dream-textures
cd dream-textures
# Install to Blender
# Install dependencies: pip install -r requirements.txt
Hardware Requirements
| Component | Minimum | Recommended |
|---|---|---|
| VRAM | 4 GB | 8-12 GB |
| GPU | CUDA 11.7+ / Apple Silicon | RTX 3080+ / M2 Max+ |
| RAM | 8 GB | 16-32 GB |
| Disk | 10 GB | 20 GB+ |
Troubleshooting
| Issue | Fix |
|---|---|
| OOM / CUDA OOM | Reduce resolution; fp16; close apps |
| Model download fails | Check internet; HF_TOKEN for gated models |
| macOS quarantine | xattr -r -d com.apple.quarantine <addon_path>/.python_dependencies |
| Slow generation | Lower steps (20-30); smaller resolution |
| Seams visible | Enable Seamless; inpaint edges |
Integration with MCP
Dream Textures is standalone - no MCP integration. Use via:
- Manual N-panel
- Python API via
execute_blender_code:
bpy.ops.dream_texture.generate(
prompt="brick wall, seamless",
seamless=True,
width=1024,
height=1024
)
- Combined workflow: GenesisCore for geometry → Dream Textures for materials
🔗 СВЯЗИ
- Прочие в категории «Референсы»: DREAM TEXTURES ADDON - DEEP DIVE, INSTALLATION GUIDE - COMPLETE SETUP, ANIMATION WORKFLOWS, BLENDER MATERIALS API REFERENCE, BLENDER MESH OPERATIONS API REFERENCE
- По темам: BLENDER PYTHON API REFERENCE (4.x), BLENDER SCRIPTING PATTERNS - SAFE & EFFECTIVE, BLENDER-MCP - FULL DOCUMENTATION
- Теги: #reference, #stable-diffusion
🎯 ЦЕЛЬ
Референс категории «Референсы», сложность —.
Связанные материалы
Подобраны по тегам и категории.