### Install Python Packages for Installer

Source: https://github.com/sakalond/stablegen/blob/main/README.md

Install the required Python packages for the installer script using pip. Ensure you have Python 3 and pip available.

```bash
pip install requests tqdm
```

--------------------------------

### Install Triton for Windows

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Installs the Triton library, specifically for Windows users. This is an optional step.

```bash
pip install triton-windows
```

--------------------------------

### Install StableGen Blender Plugin

Source: https://github.com/sakalond/stablegen/blob/main/README.md

Install the StableGen Blender addon by downloading the zip file from the releases page and installing it through Blender's Add-on preferences.

```bash
Unzip StableGen.zip and install via Blender's Add-on preferences.
```

--------------------------------

### Run the Automated Installer Script

Source: https://github.com/sakalond/stablegen/blob/main/README.md

Execute the installer script to download and place ComfyUI custom nodes and AI models. Provide your ComfyUI directory path as an argument or be prompted.

```bash
python installer.py <YourComfyUIDirectory>
```

--------------------------------

### Install Marigold Dependencies

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Installs the required Python packages for the Marigold custom node. Ensure you are in the ComfyUI-Marigold directory before running.

```bash
cd ComfyUI-Marigold
pip install -r requirements.txt
pip install "diffusers>=0.28"
```

--------------------------------

### Install ComfyUI IPAdapter Flux Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the ComfyUI IPAdapter Flux custom node repository into your custom_nodes directory. Restart ComfyUI after installation.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/Shakker-Labs/ComfyUI-IPAdapter-Flux.git
```

--------------------------------

### Install TRELLIS2 Dependencies

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Install the Python dependencies required by the ComfyUI-TRELLIS2 custom node using the provided requirements file.

```bash
pip install -r requirements.txt
```

--------------------------------

### Install comfy-env Dependency

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Install the specific version of the comfy-env package required by ComfyUI-TRELLIS2.

```bash
pip install comfy-env==0.2.0
```

--------------------------------

### Get ComfyUI VRAM Statistics

Source: https://context7.com/sakalond/stablegen/llms.txt

Retrieves free and total VRAM information from the ComfyUI server by querying the /system_stats endpoint.

```python
import urllib.request
import json

# ComfyUI server configuration (set in addon preferences)
server_address = "127.0.0.1:8188"

# Get VRAM statistics
def get_vram_stats(server_address):
    req = urllib.request.Request(
        f"http://{server_address}/system_stats", method='GET'
    )
    resp = json.loads(urllib.request.urlopen(req, timeout=10).read())
    dev = resp.get("devices", [{}])[0]
    vram_free = dev.get("vram_free", 0) / (1024 * 1024)
    vram_total = dev.get("vram_total", 0) / (1024 * 1024)
    return vram_free, vram_total
```

--------------------------------

### Check ComfyUI Server Status

Source: https://context7.com/sakalond/stablegen/llms.txt

Verifies if the ComfyUI server is reachable by making a GET request to the /system_stats endpoint.

```python
import urllib.request
import json

# ComfyUI server configuration (set in addon preferences)
server_address = "127.0.0.1:8188"

# Check server status
def check_server_alive(server_address, timeout=5):
    try:
        req = urllib.request.Request(
            f"http://{server_address}/system_stats",
            method='GET'
        )
        urllib.request.urlopen(req, timeout=timeout)
        return True
    except Exception:
        return False
```

--------------------------------

### Download SDXL Lighting LoRA

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the SDXL Lighting 8-step LoRA model, which is necessary for default presets and recommended for faster generation.

```bash
wget -O <YourComfyUIDirectory>/models/loras/sdxl_lightning_8step_lora.safetensors "https://huggingface.co/ByteDance/SDXL-Lightning/resolve/main/sdxl_lightning_8step_lora.safetensors?download=true"
```

--------------------------------

### Download IPAdapter Model

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the core IPAdapter model for SDXL. Ensure it is placed in the correct directory and named precisely as specified.

```bash
wget -O <YourComfyUIDirectory>/models/ipadapter/ip-adapter-plus_sdxl_vit-h.safetensors "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter-plus_sdxl_vit-h.safetensors?download=true"
```

--------------------------------

### Enable and Configure PBR Decomposition

Source: https://context7.com/sakalond/stablegen/llms.txt

Enables the PBR decomposition pipeline and sets various parameters for resolution, denoising, albedo extraction, tiling, and emission detection.

```python
import bpy

scene = bpy.context.scene

# Enable PBR decomposition
scene.pbr_decomposition = True

# Resolution and quality settings
scene.pbr_processing_resolution = 768    # Base resolution for processing
scene.pbr_use_native_resolution = True   # Use source image resolution
scene.pbr_denoise_steps = 4              # Denoising iterations
scene.pbr_ensemble_size = 1              # Ensemble averaging count

# Albedo extraction settings
scene.pbr_albedo_source = 'delight'      # 'delight' or 'direct'
scene.pbr_delight_strength = 1.0         # Delighting intensity
scene.pbr_albedo_auto_saturation = False # Auto saturation correction
scene.pbr_albedo_saturation_mode = 'MEDIAN'  # 'MEDIAN', 'MEAN', etc.

# Tiling for high-resolution outputs
scene.pbr_tiling = 'selective'           # 'off', 'selective', 'all', 'custom'
scene.pbr_tile_grid = 2                  # 2x2 tiling
scene.pbr_tile_superres = False          # Superresolution upscaling

# Per-map tiling (when pbr_tiling='custom')
scene.pbr_tile_albedo = True
scene.pbr_tile_material = False          # Roughness/metallic
scene.pbr_tile_normal = False
scene.pbr_tile_height = False
scene.pbr_tile_emission = False

# Emission detection settings
scene.pbr_emission_method = 'residual'   # 'residual' or 'hsv'
scene.pbr_emission_threshold = 0.2
# For HSV method:
scene.pbr_emission_saturation_min = 0.5
scene.pbr_emission_value_min = 0.85
scene.pbr_emission_bloom = 5.0
```

--------------------------------

### Download and Rename ControlNet Depth Model

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the ControlNet Depth model for SDXL. This model is crucial for StableGen presets. Rename the downloaded file to match the target filename.

```bash
wget -O <YourComfyUIDirectory>/models/controlnet/controlnet_depth_sdxl.safetensors "https://huggingface.co/xinsir/controlnet-depth-sdxl-1.0/resolve/main/diffusion_pytorch_model.safetensors?download=true"
```

--------------------------------

### Configure Sequential Mode and Texture Generation

Source: https://context7.com/sakalond/stablegen/llms.txt

Set parameters for sequential generation, including blending, IPAdapter usage, and execute texture generation. Use 'regenerate_selected' for specific viewpoints or 'stablegen_reproject' for reprojecting existing images.

```python
scene.sequential_smooth = True
scene.sequential_factor = 0.7            # Blend with previous views
scene.sequential_ipadapter = True        # Use IPAdapter for style
scene.sequential_ipadapter_mode = 'first'  # Reference first generated image

# IPAdapter settings
scene.ipadapter_weight_type = 'style'
scene.ipadapter_strength = 1.0
scene.ipadapter_start = 0.0
scene.ipadapter_end = 1.0

# Execute texture generation
bpy.ops.object.test_stable()

# For regenerating specific viewpoints only
scene.generation_mode = 'regenerate_selected'
bpy.ops.object.stablegen_regenerate()

# For reprojecting existing images with new blend settings
bpy.ops.object.stablegen_reproject()
```

--------------------------------

### Download Additional SDXL Lightning LoRA

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the SDXL Lightning 4-step LoRA model. Place it in the 'loras' directory without renaming.

```bash
wget -O <YourComfyUIDirectory>/models/loras/sdxl_lightning_4step_lora.safetensors "https://huggingface.co/ByteDance/SDXL-Lightning/resolve/main/sdxl_lightning_4step_lora.safetensors?download=true"
```

--------------------------------

### Download StableDelight Model using huggingface_hub

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Downloads the StableDelight model using the huggingface_hub Python library. Specify the correct local directory for ComfyUI models.

```python
from huggingface_hub import snapshot_download
snapshot_download('Stable-X/yoso-delight-v0-4-base', local_dir='<YourComfyUIDirectory>/models/diffusers/Stable-X--yoso-delight-v0-4-base')
```

--------------------------------

### Configure Texture Generation Parameters

Source: https://context7.com/sakalond/stablegen/llms.txt

Sets up parameters for texture generation using ComfyUI, including model architecture, generation method, sampling settings, and prompts. Configure visibility and blending parameters.

```python
import bpy

# Set up generation parameters
scene = bpy.context.scene

# Model and workflow settings
scene.model_architecture = 'sdxl'        # 'sdxl', 'flux1', 'qwen_image_edit', 'flux2_klein'
scene.generation_method = 'sequential'   # 'grid', 'separate', 'sequential', 'uv_inpaint', 'local_edit'
scene.steps = 8
scene.cfg = 1.5
scene.sampler = 'dpmpp_2s_ancestral'
scene.scheduler = 'sgm_uniform'

# Prompt configuration
scene.prompt = "a medieval stone castle with moss-covered walls"
scene.negative_prompt = "blurry, low quality, artifacts"

# Visibility and blending settings
scene.discard_factor = 90.0              # Angle threshold for visibility
scene.weight_exponent = 3.0              # Blending weight power
scene.differential_diffusion = True      # Smooth mask transitions
scene.blur_mask = True
scene.blur_mask_radius = 1
scene.grow_mask_by = 3                   # Expand mask for overlap

```

--------------------------------

### Configure TRELLIS.2 Mesh Generation

Source: https://context7.com/sakalond/stablegen/llms.txt

Set up input, generation parameters, shape, and texture settings for the TRELLIS.2 operator. This includes resolution, VRAM mode, guidance, steps, and post-processing options like decimation and background removal. Execute mesh generation with bpy.ops.object.trellis2_generate().

```python
import bpy

scene = bpy.context.scene

# Input configuration
scene.trellis2_generate_from = 'image'   # 'image' or 'prompt'
scene.trellis2_input_image = "/path/to/reference.png"
scene.trellis2_prompt = "a cute robot character"  # Used when generate_from='prompt'

# Generation parameters
scene.trellis2_resolution = '1024_cascade'  # Resolution mode
scene.trellis2_vram_mode = 'disk_offload'   # VRAM management
scene.trellis2_attn_backend = 'flash_attn'  # Attention implementation

# Shape generation settings
scene.trellis2_ss_guidance = 7.5            # Sparse structure guidance
scene.trellis2_ss_steps = 12                # Sparse structure steps
scene.trellis2_shape_guidance = 7.5         # Shape refinement guidance
scene.trellis2_shape_steps = 12             # Shape refinement steps

# Texture settings (native TRELLIS texture or external pipeline)
scene.trellis2_texture_mode = 'sdxl'        # 'none', 'native', 'sdxl', 'qwen_image_edit', 'flux2_klein'
scene.trellis2_tex_guidance = 7.5
scene.trellis2_tex_steps = 12
scene.trellis2_texture_size = 4096          # Output texture resolution

# Post-processing
scene.trellis2_decimation = 1000000         # Max polygon count
scene.trellis2_remesh = True                # Apply remeshing
scene.trellis2_bg_removal = 'auto'          # Background removal mode
scene.trellis2_import_scale = 2.0           # Import scale factor

# Camera placement for texturing phase
scene.trellis2_camera_count = 10
scene.trellis2_placement_mode = 'normal_weighted'
scene.trellis2_auto_prompts = True
scene.trellis2_exclude_bottom = True
scene.trellis2_auto_aspect = 'per_camera'

# Execute mesh generation
bpy.ops.object.trellis2_generate()
```

--------------------------------

### Download Qwen 3 4B Text Encoder

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Provides the direct download URL for the Qwen 3 4B text encoder (bf16). Save this file to the specified ComfyUI text encoders directory.

```bash
Download URL: https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/text_encoders/qwen_3_4b.safetensors?download=true
```

--------------------------------

### Configure Qwen Image Edit Settings in Blender

Source: https://context7.com/sakalond/stablegen/llms.txt

Set up various parameters for Qwen Image Edit models, including generation method, context render mode, guidance maps, custom prompts, and cleanup options. Ensure correct model architecture is selected.

```python
import bpy

scene = bpy.context.scene

# Qwen-specific settings
scene.model_architecture = 'qwen_image_edit'
scene.qwen_generation_method = 'generate'  # 'generate', 'local_edit', 'refine'

# Context render mode - how previous textures are incorporated
scene.qwen_context_render_mode = 'REPLACE_STYLE'  # 'NONE', 'REPLACE_STYLE', 'ADDITIONAL'

# Guidance map settings
scene.qwen_guidance_map_type = 'depth'
scene.qwen_guidance_fallback_color = (1.0, 0.0, 1.0)  # Magenta for untextured areas
scene.qwen_guidance_background_color = (1.0, 0.0, 1.0)

# Custom prompts (when qwen_use_custom_prompts=True)
scene.qwen_use_custom_prompts = False
scene.qwen_custom_prompt_initial = "Change the format of image 1 to '{main_prompt}'"
scene.qwen_custom_prompt_seq_none = "Change and transfer the format of '{main_prompt}' in image 1 to the style from image 2"
scene.qwen_custom_prompt_seq_replace = "Change and transfer the format of image 1 to '{main_prompt}'. Replace all solid magenta areas in image 2."

# Context cleanup - remove residual fallback colors
scene.qwen_context_cleanup = False
scene.qwen_context_cleanup_hue_tolerance = 0.0
scene.qwen_context_cleanup_value_adjust = 0.0
scene.qwen_context_fallback_dilation = 1

# Other Qwen options
scene.qwen_prompt_gray_background = True
scene.qwen_rescale_alignment = True      # Align to 112-pixel boundaries
scene.qwen_voronoi_mode = False          # Hard camera segmentation
scene.qwen_use_external_style_image = False
scene.qwen_external_style_image = ""     # Path to style reference
```

--------------------------------

### Add Cameras with Fan from Camera Placement

Source: https://context7.com/sakalond/stablegen/llms.txt

Places cameras in a fan arc around the active camera's position. Specify the number of cameras and the fan angle.

```python
import bpy

# Fan from Camera - arc around active camera position
bpy.ops.object.add_cameras(
    placement_mode='fan_from_camera',
    num_cameras=5,
    fan_angle=90.0                       # 90 degree arc spread
)
```

--------------------------------

### Apply StableGen Presets

Source: https://context7.com/sakalond/stablegen/llms.txt

Utilize the preset system to quickly configure generation parameters for various use cases. Assign a preset name to scene.stablegen_preset to apply its settings.

```python
import bpy
from stablegen.ui.presets import PRESETS, apply_preset

# Available presets include:
# "DEFAULT" - General purpose SDXL generation
# "CHARACTERS" - Optimized for character texturing with IPAdapter
# "ARCHITECTURE" - High weight exponent for flat surfaces
# "QUICK DRAFT" - Fast 4-step generation for previews
# "UV INPAINTING" - For filling untextured UV regions
# "QWEN PRECISE" - Qwen Image Edit with context renders
# "QWEN SAFE" - Qwen with previous view reference for consistency
# "QWEN VORONOI" - Hard camera segmentation during generation
# "KLEIN PRECISE/SAFE/ALT" - FLUX.2 Klein variants
# "LOCAL REFINE" - SDXL local edit for specific areas
# "LOCAL EDIT (QWEN)" - Targeted edits with Qwen
# "DEFAULT (MESH + TEXTURE)" - Full TRELLIS.2 + SDXL pipeline
# "CHARACTERS (MESH + TEXTURE)" - TRELLIS.2 optimized for characters
# "QWEN PRECISE (MESH + TEXTURE)" - TRELLIS.2 + Qwen pipeline

# Apply a preset
scene = bpy.context.scene
scene.stablegen_preset = "CHARACTERS"

# Presets contain settings like:
preset = PRESETS["CHARACTERS"]
# preset["model_architecture"] = "sdxl"
# preset["generation_method"] = "sequential"
# preset["steps"] = 8
# preset["cfg"] = 1.5
# preset["sequential_ipadapter"] = True
# preset["discard_factor"] = 80.0
# preset["weight_exponent"] = 3.0
# preset["controlnet_units"] = [...] 
# preset["lora_units"] = [...]
```

--------------------------------

### Import bpy, bmesh, numpy, and BVHTree

Source: https://context7.com/sakalond/stablegen/llms.txt

Imports necessary libraries for Blender scripting, including bmesh for mesh manipulation, numpy for numerical operations, and BVHTree for spatial queries.

```python
import bpy
import bmesh
import numpy as np
from mathutils.bvhtree import BVHTree
```

--------------------------------

### Download FLUX.2 Klein Diffusion Model

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Provides the direct download URL for the FLUX.2 Klein base diffusion model (FP8). Save this file to the specified ComfyUI models directory.

```bash
Download URL: https://huggingface.co/black-forest-labs/FLUX.2-klein-base-4b-fp8/resolve/main/flux-2-klein-base-4b-fp8.safetensors?download=true
```

--------------------------------

### Download FLUX.2 VAE

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Provides the direct download URL for the FLUX.2 VAE. Save this file to the specified ComfyUI VAE directory.

```bash
Download URL: https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors?download=true
```

--------------------------------

### Download and Rename IPAdapter CLIP Vision Model 2

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the second CLIP Vision model required for IPAdapter. Rename the downloaded file to the specified target filename.

```bash
wget -O <YourComfyUIDirectory>/models/clip_vision/CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/image_encoder/model.safetensors"
```

--------------------------------

### Configure LoRA Units

Source: https://context7.com/sakalond/stablegen/llms.txt

Sets up LoRA units with model names and strengths for both the model and CLIP components.

```python
lora_config = [
    {
        'model_name': 'sdxl_lightning_8step_lora.safetensors',
        'model_strength': 1.0,
        'clip_strength': 1.0
    }
]
```

--------------------------------

### Enable Online Access in Blender

Source: https://github.com/sakalond/stablegen/blob/main/README.md

Enable online access in Blender's system preferences. This is a requirement for some add-ons, including StableGen, even if they primarily operate locally.

```bash
Edit -> Preferences -> System -> Network -> Enable Online Access
```

--------------------------------

### Download and Rename IPAdapter CLIP Vision Model 1

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Download the first CLIP Vision model required for IPAdapter. You must rename the downloaded file to match the target filename.

```bash
wget -O <YourComfyUIDirectory>/models/clip_vision/CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors "https://huggingface.co/h94/IP-Adapter/resolve/main/models/image_encoder/model.safetensors"
```

--------------------------------

### Configure ControlNet Units

Source: https://context7.com/sakalond/stablegen/llms.txt

Add and configure multiple ControlNet units for geometric guidance. Each unit can specify control type, model, strength, and timing.

```python
import bpy

scene = bpy.context.scene

# ControlNet units are configured as a list of dictionaries

```

--------------------------------

### Add Cameras with Normal-Weighted Placement

Source: https://context7.com/sakalond/stablegen/llms.txt

Places cameras around an object using normal-weighted auto placement. Configure coverage targets, occlusion modes, and interactive review.

```python
import bpy

# Place 8 cameras using normal-weighted auto placement
bpy.ops.object.add_cameras(
    placement_mode='normal_weighted',    # Auto placement based on mesh normals
    num_cameras=8,                       # Number of cameras to place
    center_type='object',                # Center on active object
    purge_others=True,                   # Remove existing cameras first
    consider_existing=True,              # Avoid duplicating coverage
    auto_aspect='per_camera',            # Compute optimal aspect per camera
    exclude_bottom=True,                 # Skip bottom-facing geometry
    exclude_bottom_angle=1.5533,         # 89 degrees threshold
    coverage_target=0.95,                # Target 95% surface coverage
    max_auto_cameras=12,                 # Maximum for greedy mode
    auto_prompts=True,                   # Generate view prompts (front, rear, etc.)
    occlusion_mode='none',               # Fast back-face culling only
    review_placement=True                # Interactive fly-through review
)
```

--------------------------------

### Clone Nunchaku Custom Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the Nunchaku repository into your ComfyUI custom nodes directory to enable optimized inference for Qwen models.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/nunchaku-tech/ComfyUI-nunchaku.git
```

--------------------------------

### Clone Qwen Image LoRA Loader

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the Qwen Image LoRA Loader repository to manage LoRAs for Qwen models within ComfyUI.

```bash
git clone https://github.com/ussoewwin/ComfyUI-QwenImageLoraLoader.git
```

--------------------------------

### Flush ComfyUI VRAM and Unload Models

Source: https://context7.com/sakalond/stablegen/llms.txt

Sends a POST request to the /free endpoint on the ComfyUI server to unload models and free up VRAM.

```python
import urllib.request
import json

# ComfyUI server configuration (set in addon preferences)
server_address = "127.0.0.1:8188"

# Flush VRAM and unload models
def flush_vram(server_address):
    data = json.dumps({"unload_models": True, "free_memory": True}).encode('utf-8')
    req = urllib.request.Request(
        f"http://{server_address}/free", data=data,
        headers={'Content-Type': 'application/json'}, method='POST'
    )
    urllib.request.urlopen(req, timeout=10)
```

--------------------------------

### Clone ComfyUI-TRELLIS2 Repository

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the ComfyUI-TRELLIS2 repository at a specific commit to enable 3D mesh generation from images or text prompts.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/PozzettiAndrea/ComfyUI-TRELLIS2.git
cd ComfyUI-TRELLIS2
git checkout 6b0b2148f45bbafa0b86bfd25c63602b63a7aae0
```

--------------------------------

### Clone ComfyUI-Marigold Custom Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the ComfyUI-Marigold repository to enable the decomposition of color textures into PBR material maps.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/kijai/ComfyUI-Marigold.git
```

--------------------------------

### Clone StableDelight Custom Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clones the ComfyUI_StableDelight_ll custom node repository into your ComfyUI custom nodes directory. Navigate to the directory first.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/lldacing/ComfyUI_StableDelight_ll.git
```

--------------------------------

### Upload Image to ComfyUI

Source: https://context7.com/sakalond/stablegen/llms.txt

Provides a placeholder for uploading an image to the ComfyUI server via the /upload/image endpoint. Requires multipart form data.

```python
import urllib.request
import json

# ComfyUI server configuration (set in addon preferences)
server_address = "127.0.0.1:8188"

# Upload image to ComfyUI
def upload_image(server_address, image_path, subfolder=""):
    with open(image_path, 'rb') as f:
        image_data = f.read()
    # ... multipart form upload to /upload/image endpoint
```

--------------------------------

### Add Cameras with Greedy Coverage

Source: https://context7.com/sakalond/stablegen/llms.txt

Uses greedy coverage mode to automatically determine the number of cameras needed for a target coverage. Supports full matrix occlusion and elevation clamping.

```python
import bpy

# Greedy coverage mode - automatically determines camera count
bpy.ops.object.add_cameras(
    placement_mode='greedy_coverage',
    coverage_target=0.98,
    max_auto_cameras=20,
    occlusion_mode='full_matrix',        # BVH-validated visibility
    clamp_elevation=True,                # Avoid extreme angles
    max_elevation_angle=1.2217           # 70 degrees max
)
```

--------------------------------

### Configure ControlNet Units and Canny Thresholds

Source: https://context7.com/sakalond/stablegen/llms.txt

Defines ControlNet units with specific model names, strengths, and start/end percentages. Also sets low and high thresholds for Canny edge detection.

```python
controlnet_config = [
    {
        'unit_type': 'depth',
        'model_name': 'controlnet_depth_sdxl.safetensors',
        'strength': 0.5,
        'start_percent': 0.0,
        'end_percent': 1.0,
        'is_union': False,
        'use_union_type': True
    },
    {
        'unit_type': 'canny',
        'model_name': 'controlnet_canny_sdxl.safetensors',
        'strength': 0.3,
        'start_percent': 0.0,
        'end_percent': 0.8
    }
]

# Canny edge detection thresholds
scene.canny_threshold_low = 0
scene.canny_threshold_high = 80
```

--------------------------------

### Clone ComfyUI IPAdapter Plus Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Navigate to your ComfyUI custom nodes directory and clone the IPAdapter Plus repository. This node is essential for IPAdapter support in StableGen.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/cubiq/ComfyUI_IPAdapter_plus.git
```

--------------------------------

### Clone ComfyUI-GGUF Custom Node

Source: https://github.com/sakalond/stablegen/blob/main/docs/MANUAL_INSTALLATION.md

Clone the ComfyUI-GGUF repository to enable support for GGUF (quantized) models. Navigate to your ComfyUI custom nodes directory before running this command.

```bash
cd <YourComfyUIDirectory>/custom_nodes/
git clone https://github.com/city96/ComfyUI-GGUF.git
```

--------------------------------

### Cancel ComfyUI Generation

Source: https://context7.com/sakalond/stablegen/llms.txt

Sends a request to the /interrupt endpoint on the ComfyUI server to cancel a currently running generation job, identified by client_id.

```python
import urllib.request
import json

# ComfyUI server configuration (set in addon preferences)
server_address = "127.0.0.1:8188"

# Cancel running generation
def cancel_generation(server_address, client_id):
    data = json.dumps({"client_id": client_id}).encode('utf-8')
    req = urllib.request.Request(
        f"http://{server_address}/interrupt", data=data
    )
    urllib.request.urlopen(req)
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.