🚀 ComfyUI Auto-Installer — v5 (Python Rewrite)
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: irm https://get.umeai.art/migrate.ps1 | iex
⚡ Quick Start (One-Liner)
Windows (PowerShell):
irm https://get.umeai.art/comfyui.ps1 | iexLinux / macOS:
curl -fsSL https://get.umeai.art/comfyui.sh | sh
Only requires Git — everything else (Python, uv, dependencies) is handled automatically.
✨ What's New in v5
Full Python rewrite — no more PowerShell dependency
Cross-platform — Windows, Linux, macOS, and Docker
TUI Manager — interactive terminal UI to launch, update, download models, and configure settings
VRAM-aware model catalog — 7 model families with quantization recommendations based on your GPU
GPU auto-detection — NVIDIA (CUDA 13.0/12.8), AMD (ROCm/DirectML), Apple Silicon (MPS)
SageAttention 2 + 3 — pre-compiled wheels including RTX 50XX Blackwell support
One-click update — update ComfyUI, all nodes, and dependencies with a single command
Model security scanner — detects malicious pickle code in .ckpt/.pt files
Junction architecture — models and outputs persist independently from ComfyUI updates
Docker ready — 4 image variants including a cloud version with JupyterLab for RunPod
📋 Prerequisites
Git
GPU: NVIDIA (CUDA 12.x+), AMD (Radeon RX 6000+), or Apple Silicon (M1+)
Internet connection
Note: Python is automatically installed via uv if not present. No manual Python setup required.
🎨 Model Catalog (7 Families)
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:
FLUX (Image): Dev, Fill
Z-IMAGE (Image): Turbo
WAN 2.1 (Video): T2V, I2V 480p
WAN 2.2 (Video): I2V, Fun Inpaint, Fun Camera
HiDream (Image): Dev
QWEN (Image Edit): Image Edit
LTX-2 (Video + Audio): Dev
🧩 34 Custom Nodes Included
Additive manifest — never removes user-installed nodes.
Core (always installed): ComfyUI-Manager
UmeAiRT Tier: ComfyUI-UmeAiRT-Sync, ComfyUI-UmeAiRT-Toolkit, ComfyUI-Crystools, ComfyUI-nunchaku
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
⚙️ GPU Optimizations (Auto-Installed)
PyTorch 2.10: CUDA 13.0/12.8, ROCm 7.1, DirectML, MPS
xformers: Memory-efficient attention
Triton: triton-windows / triton (Linux)
SageAttention 2: Unified ABI3 wheels (Windows), per-arch SM80–SM100 (Linux)
SageAttention 3: RTX 50XX Blackwell native (Windows + Linux)
FlashAttention: Linux + NVIDIA only
Nunchaku & InsightFace: Pre-compiled wheels
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.
🐳 Docker Support
Requires Docker and an NVIDIA GPU: docker run --gpus all -p 8188:8188 -v comfyui-data:/data registry.gitlab.com/umeairt-studio/comfyui-auto_installer-python:latest
latest: ~4 GB — Ready to go with pre-installed PyTorch
latest-cloud: ~4.5 GB — + JupyterLab for RunPod / cloud
latest-lite: ~2 GB — Minimal (installs PyTorch on first run)
latest-lite-cloud: ~2 GB — Lite + JupyterLab
🔒 Security
No external script execution — all logic is internalized
Secure subprocess calls — no shell=True
HTTPS only — all URLs validated
SHA-256 integrity checks on all model downloads
Pickle model scanner — detects malicious code in .ckpt/.pt files
Zip-slip prevention on archive extraction
CI runs Bandit + pip-audit on every push
📂 Post-Installation
Three launcher scripts are generated:
UmeAiRT-Start-ComfyUI: Launch (Performance mode + SageAttention)
UmeAiRT-Start-ComfyUI_LowVRAM: Launch with --lowvram --fp8 for ≤8 GB VRAM
UmeAiRT-Manager: TUI manager (update, download, reinstall, settings)
🔗 Links
Source code: GitLab (https://gitlab.com/UmeAiRT-Studio/ComfyUI-Auto_installer-Python)
Mirror: Codeberg (https://codeberg.org/UmeAiRT)
Ecosystem: UmeAiRT Studio (https://umeai.art)
Description
What's new? :
ComfyUi 0.3.30,
pytorch version: 2.7.0+cu128,
xformers version: 0.0.30,
support for 50X0 cards,
SageAttention, triton and Apex included.
FAQ
Comments (43)
perfect I just got a 5080 friday it has been hell to update videomodels for it, thank you very much, I will try it right away
the install worked...I think, I believe it overwrote my existing comfyUI, many dates of today in it.
but I still have problems getting the nodes in your workflows to work, updated or otherwise, wanimagetovideo, cfgzerostar, cliploadermultigpu, unetloaderggufdistorchmultigpu are all red outlined
You have to use it in a new folder, not for update an existing installation.
@UmeAiRT Aha ok, and if I put it in another folder will it install there?
@serget2 As long as you don't run it as administrator (that moves everything to system32 ^^') it installs in the folder in which it is
@UmeAiRT yeah I was looking everywhere but it installed inside the folder, my brains are fried form trying to solve this for the past 2 days now, I am currently moving the folder to a place where I can find it and then I will test it
THE MODEL LOADER CAN'T FIND ANY MODELS, sorry caps lock was on, I have the latest image to video but the unet name is undefined, I tried manually moving the models to unet and back to diffuser models and tried the load gguf model it doesn't work, so where do I put the models?
@serget2 for wan model it's in diffusion_models but you have try the model downloader script downloaded during instalation?
@UmeAiRT I did so it is strange I asked it for all the models and I have the =m in diffusion models but it doesn't seem to find it. Ima try again
Would the "LOWVRAM" option be for those who have less than 12GB of video? Can I use the normal executable even if I have a card with 8GB of video?
This is an option offered by ComfyUI, the easiest thing is to try it to see if it works better for you.
@UmeAiRT Thanks!
After installation I noticed that 4 more .bat files were created, should I run them?
Both "model downloader" scripts ask questions like during installation to download additional models. The missing node downloader allows you to remove and reinstall additional nodes in case of problems; it should not be used when everything is working.
@UmeAiRT So there is no need to use them if everything is ok, I understand, thank you very much!
i've been using A1111 for a couple years but I know absolutely nothing about Comfy or WAN, but I want to get started. Is your script the right place to start? Also, two specific questions:
- Do you know how much drive space is needed in total for this starting kit?
- If I want to use SDXL instead can I edit the Flux installs out of the script without breaking anything?
How critical is this message?
Start-Process : The command cannot be executed due to the following error: The specified file cannot be found.
Line:1 sign:1
+ Start-Process winget -WindowStyle Hidden -ArgumentList 'install','--i ...
+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ CategoryInfo : InvalidOperation: (:) [Start-Process], InvalidOperationException
+ FullyQualifiedErrorId : InvalidOperationException,Microsoft.PowerShell.Commands.StartProcessCommand
- Python include/libs
- NVIDIA Apex
- Triton
- xformers
- SageAttention
Would you like to download WAN models?
Enter your choice (Y or N) and press Enter:
It seems that this command did not work for you:
winget install --id Microsoft.VisualStudio.2022.BuildTools -e --source winget --override "--quiet --wait --norestart --add Microsoft.VisualStudio.Component.VC.Tools.x86.x64 --add Microsoft.VisualStudio.Component.Windows10SDK.20348"
So you need to manually install Visual Studio Tool to use all workflow functions like TorchCompile.
For a very long time, I was tormented with the installation, I had to update Windows in order for Visual Studio to work correctly, then I had to register the PATH to winget.
tell me, for 12 GB of video memory, which model should I use? bf16, fp16, fp8
For 12 GB i recommend GGUF Q4_K_S version
Then skip the installation here, right?
Do you want to upload WAN models?
A) bf16
B) fp16
C) fp8
D) All
E) No
Enter the values you selected (A, B, C, DorE) and press Enter:
@smolusha say no to this and the next question is for gguf models
Everything has been successfully established. Tell me which batch file to run for a 12 GB graphics card
run_cpu
run_nvidia_gpu
run_nvidia_gpu_fast_fp16_accumulation
run_nvidia_gpu-LOWVRAM
run_nvidia_gpu-sageattention
run_nvidia_gpu-sageattention
Thank you so much. This is the best build for wan2.1. Thank you for answering the questions. Anyone who reads the comments, feel free to put this ComfyUI
when i try to run for generate VIDEO i stuck at
Requested to load WAN21
loaded partially 18970.114999999998 18970.114868164062 0
0% | 0/10 [00:00<?, ?it/s]
for a long time
Is it normal? when i generate it's always Requested to load WAN21 and i must wait for almost 1 hour
when i try in GGUF it's still fast but when i try in Wan/base it use so long time
It looks like you don't have enough VRAM to use the base model. This is easily seen because the model is partially loaded.
@UmeAiRT I use 4090 . Or i set wrong setting?
@shadowwave01213 4090 have 24GB of VRAM, if you use fp16 model (+30GB) you have memory overflow
@UmeAiRT Can you suggest What the best solution if i want to IMG to VIDEO or TXT to VIDEO And sorry for annoy question i'm newbie and i ask gpt then it send me here
@shadowwave01213 For your graphics card I would recommend the fp8 version in "base" or Q8 in "GGUF". To put it simply, avoid using a model that is larger than your VRAM.
@UmeAiRT So i should Open UmeAiRT-WAN2.1-Model_downloader and choose fp8 version right?
@shadowwave01213 Yes i think it's the beter model for 4090, i dont have one so i cant test myself
I'm wondering if it's realistic to run Hunyuan-video with its models on your workflows?
I have not found anything similar and convenient on this site as your workflow.
Is this using Sage Attention 2 or 1? Thanks for the installers man, really appreciated
Thank you so much, when i run wan2.1, my 3090 vram fills quickly and i get OOM error no matter which model of Wan i use(1.3,480p,720p,tensor,gguf),i don't know what else should i do
the only way i was able to run Wan on my rig was just downloading Comfyui Portable and i did run with just torch attention...
when i'll try to run the UmeAirt workflow it shows - The safetensors file is corrupt or invalid. Make sure this is actually a safetensors file and not a ckpt or pt or other filetype.
Model files are often large, and sometimes the download glitches. You then have to delete the corrupted file and download it manually. All the links for each file are in my workflow description.
For a 50 series GPU should we be installing nightly pytorch instead?