<h1 id="comfyui-auto-installer-v5-(python-rewrite)">🚀 ComfyUI Auto-Installer — v5 (Python Rewrite)</h1>Version 5 is a full rewrite from the ground up in Python, replacing all the PowerShell scripts from previous versions. It's cross-platform, faster, smarter, and now ships with a TUI manager, Docker images, and GPU-optimized inference out of the box.If you are upgrading from the PowerShell version (v4.x), a one-command migration preserves all your models, outputs, and custom nodes: <code>irm https://get.umeai.art/migrate.ps1 | iex</code><h2 id="quick-start-(one-liner)">⚡ Quick Start (One-Liner)</h2><ul><li>Windows (PowerShell): <code>irm https://get.umeai.art/comfyui.ps1 | iex</code></li><li>Linux / macOS: <code>curl -fsSL https://get.umeai.art/comfyui.sh | sh</code></li></ul>Only requires Git — everything else (Python, uv, dependencies) is handled automatically.<h2 id="what's-new-in-v5">✨ What's New in v5</h2><ul><li>Full Python rewrite — no more PowerShell dependency</li><li>Cross-platform — Windows, Linux, macOS, and Docker</li><li>TUI Manager — interactive terminal UI to launch, update, download models, and configure settings</li><li>VRAM-aware model catalog — 7 model families with quantization recommendations based on your GPU</li><li>GPU auto-detection — NVIDIA (CUDA 13.0/12.8), AMD (ROCm/DirectML), Apple Silicon (MPS)</li><li>SageAttention 2 + 3 — pre-compiled wheels including RTX 50XX Blackwell support</li><li>One-click update — update ComfyUI, all nodes, and dependencies with a single command</li><li>Model security scanner — detects malicious pickle code in .ckpt/.pt files</li><li>Junction architecture — models and outputs persist independently from ComfyUI updates</li><li>Docker ready — 4 image variants including a cloud version with JupyterLab for RunPod</li></ul><h2 id="prerequisites">📋 Prerequisites</h2><ul><li>Git</li><li>GPU: NVIDIA (CUDA 12.x+), AMD (Radeon RX 6000+), or Apple Silicon (M1+)</li><li>Internet connection</li><li>Note: Python is automatically installed via uv if not present. No manual Python setup required.</li></ul><h3 id="model-catalog-(7-families)">🎨 Model Catalog (7 Families)</h3>Interactive model downloader with VRAM-based recommendations (★ markers) and SHA-256 integrity checks. Each bundle offers multiple quantization variants (fp16, fp8, GGUF Q3→Q8). Downloads are accelerated via aria2c with HuggingFace + ModelScope fallback:<ul><li>FLUX (Image): Dev, Fill</li><li>Z-IMAGE (Image): Turbo</li><li>WAN 2.1 (Video): T2V, I2V 480p</li><li>WAN 2.2 (Video): I2V, Fun Inpaint, Fun Camera</li><li>HiDream (Image): Dev</li><li>QWEN (Image Edit): Image Edit</li><li>LTX-2 (Video + Audio): Dev</li></ul><h3 id="34-custom-nodes-included">🧩 34 Custom Nodes Included</h3>Additive manifest — never removes user-installed nodes.<ul><li>Core (always installed): ComfyUI-Manager</li><li>UmeAiRT Tier: ComfyUI-UmeAiRT-Sync, ComfyUI-UmeAiRT-Toolkit, ComfyUI-Crystools, ComfyUI-nunchaku</li><li>Full Tier (all of the above +): ComfyUI-Impact-Pack, ComfyUI-Impact-Subpack, ComfyUI-GGUF, ComfyUI-mxToolkit, ComfyUI-Custom-Scripts, ComfyUI-KJNodes, ComfyUI-WanVideoWrapper, ComfyUI-VideoHelperSuite, ComfyUI-Frame-Interpolation, rgthree-comfy, ComfyUI-Easy-Use, ComfyUI-HunyuanVideoMultiLora, ComfyUI-Florence2, ComfyUI-MultiGPU, ComfyUI-WanStartEndFramesNative, ComfyUI-Image-Saver, ComfyUI_UltimateSDUpscale, comfyui_controlnet_aux, x-flux-comfyui, ComfyUI-Detail-Daemon, wlsh_nodes, ComfyUI_essentials, ComfyUI-wanBlockswap, Derfuu_ComfyUI_ModdedNodes, ComfyUI_LayerStyle, ComfyUI-Upscaler-Tensorrt, comfyui-vrgamedevgirl, comfyui-int-and-float, was-node-suite-comfyui</li></ul><h3 id="gpu-optimizations-(auto-installed)">⚙️ GPU Optimizations (Auto-Installed)</h3><ul><li>PyTorch 2.10: CUDA 13.0/12.8, ROCm 7.1, DirectML, MPS</li><li>xformers: Memory-efficient attention</li><li>Triton: triton-windows / triton (Linux)</li><li>SageAttention 2: Unified ABI3 wheels (Windows), per-arch SM80–SM100 (Linux)</li><li>SageAttention 3: RTX 50XX Blackwell native (Windows + Linux)</li><li>FlashAttention: Linux + NVIDIA only</li><li>Nunchaku &amp; InsightFace: Pre-compiled wheels</li><li>Additional Python packages auto-installed: facexlib, onnxruntime-gpu, nvidia-ml-py, cupy-cuda13x, imageio-ffmpeg, hf_xet, cython, rotary_embedding_torch, blend_modes, segment_anything, gguf, and more.</li></ul><h3 id="docker-support">🐳 Docker Support</h3>Requires Docker and an NVIDIA GPU: <code>docker run --gpus all -p 8188:8188 -v comfyui-data:/data registry.gitlab.com/umeairt-studio/comfyui-auto_installer-python:latest</code><ul><li>latest: ~4 GB — Ready to go with pre-installed PyTorch</li><li>latest-cloud: ~4.5 GB — + JupyterLab for RunPod / cloud</li><li>latest-lite: ~2 GB — Minimal (installs PyTorch on first run)</li><li>latest-lite-cloud: ~2 GB — Lite + JupyterLab</li></ul><h3 id="security">🔒 Security</h3><ul><li>No external script execution — all logic is internalized</li><li>Secure subprocess calls — no shell=True</li><li>HTTPS only — all URLs validated</li><li>SHA-256 integrity checks on all model downloads</li><li>Pickle model scanner — detects malicious code in .ckpt/.pt files</li><li>Zip-slip prevention on archive extraction</li><li>CI runs Bandit + pip-audit on every push</li></ul><h3 id="post-installation">📂 Post-Installation</h3>Three launcher scripts are generated:<ul><li>UmeAiRT-Start-ComfyUI: Launch (Performance mode + SageAttention)</li><li>UmeAiRT-Start-ComfyUI_LowVRAM: Launch with --lowvram --fp8 for ≤8 GB VRAM</li><li>UmeAiRT-Manager: TUI manager (update, download, reinstall, settings)</li></ul><h3 id="links">🔗 Links</h3><ul><li>Source code: GitLab (<a target="_blank" rel="ugc" href="https://gitlab.com/UmeAiRT-Studio/ComfyUI-Auto_installer-Python">https://gitlab.com/UmeAiRT-Studio/ComfyUI-Auto_installer-Python</a>)</li><li>Mirror: Codeberg (<a target="_blank" rel="ugc" href="https://codeberg.org/UmeAiRT">https://codeberg.org/UmeAiRT</a>)</li><li>Ecosystem: UmeAiRT Studio (<a target="_blank" rel="ugc" href="https://umeai.art">https://umeai.art</a>)</li></ul>

【TOOL】ComfyUI Installer

Added an option to install Pytorch 2.13 with Kitchen attention instead of Pytorch 2.10 and Sage-attention

Files with SHA256: d212f85a67e6b4174eefabd10fce5c6847af2211123b2987d2d2d733fd3b88ed