🚀 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
Update to :
python 3.13,
torch 2.10.0,
nunchaku 1.2.1.
New triton and sage-attention instalation method
FAQ
Comments (30)
lol I also vibe coded an install script, but mine I also made sure to include safe guards like ensuring it only mounts localhost, prevents other ports, auto configs firewalld, installs it to application launcher, sets up venv, gets all dependencies, and then also makes sure on the launch script thats created, to do checks for security each time, checking any custom nodes for susp calls like curl, etc. (for malware).
Idk if I would release one for others without at least including proper protocols. Not to mention, if you just made if for your distro, that can have issues as each install can be different, so a "universal" install baseline would be more ideal for a "release"
If you just install it base, MANY people leave their instance WIDE open to the internet.
Cool story bro
@navyguy757 where urs at?
Pics or it didn't happen, anonymous non-contributing shit talker
can I run this on external HDD? if I cancel it, will it ask me again what's to install or will it continue from the last progress?
usually it is recommended to run ai stuff on SSD at least sata if not nvme m.2 for faster generations but HDD works fine but will be slow compared to an ssd
I think i found a possible bug, its a bit frustrating xD. i updated with the faulty v43. and it broke it bad enough i had to do a clean install. v43 still wasn't working for me so i just reinstalled v42 (it works fine). but i noticed its not reinstalling dependancies and requirements for nodes that already have an existing folder. so those nodes break and are deemed "not installed" as it skips over them in both v42 and v43. maybe for clean install you can add an option to remove existing folders for things like nodes and models for a complete wipe and reinstall?
I ran your installer, everything worked. i can start comfy, but every node has a red border. It seems they are all installed and updated, but nothing works. am i missing something?
can i be that some comfy updates ruined the workflow?
Install Required: Text Find and Replace
Install Required: MagCache
Install Required: LayerUtility: ImageScaleByAspectRatio V2
i install them, but i only get error messages. updated, restarted, removed all nodes and installed them again. tried dfferent versions. nothing works
You might try what I just did. The latest installer corrupts basic functionality in Comfy-Core. I opened the Manager interface and selected Update ComfyUI and that seemed to solve the issue.
@BRai4now sadly No result
Hi again :)
Loving your updates, thanks for them.
However, I have just updated from 4.1 to 4.3 and some workflows are working and on my main one that I'm in the middle of a project I'm getting this error:
* KSampler 3:
- Return type mismatch between linked nodes: scheduler, received_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent']) mismatch input_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent', 'beta57'])
The K Sampler and S2VExtend boxes are red 🤔
I hope you can help, thank you.
not working for me. get this message after installing then starting Comfy:
'chcp' is not recognized as an internal or external command,
operable program or batch file.
[INFO] Checking installation type...
[INFO] Activating venv environment...
[ERROR] Failed to activate venv environment.
Press any key to continue . . .
At first I wasn't able to start Comfy after running the 4.3 install, though no errors were detected. Turned out it just didn't like my current driver version. Updated that and Comfy started right up.
However, now it doesn't know from the stock, base positive and negative input nodes. They're red.
Node 'Positive' has no class_type. The workflow may be corrupted or a custom node is missing.
...this is from trying to run your txt2v wan2.2 but they show up with red borders just launching Comfy with the stock workflow.
Got it working. The install buggers Comfy-core somehow. I took my chances running the option in manager to update Comfy and that seemed to fix it. Interesting though, doing a second install (because I do video and image gen in two separate installs on separate drives), the second needed no update after the install. Strangeness.
One unexpected "feature" seems to also be that your UI scale comes across in the install, making the ComfyUI UI almost unworkable on a 1080 height monitor. It was weird because it wasn't my browser scale, just launching the fresh new install while my older install had a normal looking UI. I'd never have thought to even try CTRL-Mousewheel because I knew I was at 100% haha. I was going crazy looking for scale controls in settings menus, lol.
I get an error where the script closes PowerShell after the models and doesn't seem to complete.
I had to downgrade transformer to 5.3.0; with 5.4.0 the florence2 nodes were not working
Florence nodes have trouble keeping up, I have been facing the same issues but decided to work without florence untill they do
Just an heads up to you guys doing a clean install.
After 18 months of working on these workflows I needed to wipe my PC and re-install Windows (long story I won't bore you with), anyway, when I was trying to set this up again I kept getting an error about Python 3.12 being missing, even though it demands 3.13 for the install and that is what gets installed.
After messing around I worked out you MIST install 312 and 313 to this to work, so there must be something in one of the scripts that was missed out.
You can find all the versions here:
Python Releases for Windows | Python.org
I hope this helps some of you :)
I did get another error right at the end but I have forgotten what it was and it does not seem to be stopping my system running the workflow.
This is the last error, it is definitely an issue where it needs 312, but then only expects 313 to be installed:
hint: The wheel is compatible with CPython 3.13 cp313), but you're using CPython 3.12 cp312)
- FATAL ERROR executing: uv pip install --python I:/UMIEART WAN 5090/scripts/venv/Scripts/python.exe I:/UMIEART WAN 5090/scripts/nunchaku-1.2.1+cu13.0torch2.10-cp313-cp313-win_amd64.whl. Error: Command execution failed. Check logs.
A fatal error occurred. Press Enter to exit.:
Currently, due to a shadowban by GitHub, all my online tools have been suspended.
@UmeAiRT Any idea when it will be online again?
@hyramyy414 I'm waiting for a response from support; apparently, I'm not human.
@UmeAiRT good luck getting it sorted, I love your workflows :)
@UmeAiRT any response? So unfair man, hope you'll get it sorted eventually
@rozowy30cm69
Hello,
GitHub is meant to be used as a collaborating and hosting tool for software developers. We're unable to see activity on your account that indicates that it's being used for the intended purpose. For this reason we will not be removing the restrictions from this account at this time.
If you believe we should have reached a different decision, please review our process for appeals and submit additional information via the instructions found here:
GitHub Appeal and Reinstatement
Note: To avoid seeing this message again, please be sure to select yes when asked if you have previously contacted GitHub about this claim when opening a new support ticket. Any replies to this email will not be seen by our team, you will need to open a new support ticket.
Best regards,
GitHub Support
@UmeAiRT Asshats, sorry to see this dude!
Wondering what could possibly trigger a reason to shadowban you.
But, why don't you just adapt it to work solely from huggingface instead? I mean, it supports git.
To avoid being banned from huggingface for misuse of their services (intended for models and datasets, not code)