🛠️ [TOOL] DaSiWa ComfyUI Installer
Details
Download Files (1)
About this version
Model description
DaSiWa ComfyUI Installer
One binary. Walk away. Come back to a working ComfyUI.
A professional-grade installer for [ComfyUI](https://github.com/comfyanonymous/ComfyUI) built on a zero-conflict, fully isolated architecture. No Python knowledge required. No admin rights needed. No system files touched.

Unified One-Page Installer
The installer now presents the full setup on a single local page. Folder selection, install mode, GPU and CUDA target, SageAttention, FFmpeg, optional downloads, and the final install plan are all visible at once, with no separate pages or script chain to follow.
That unified layout is deliberate:
Choose the ComfyUI folder once.
Pick the install mode once.
Review hardware and version overrides in the same view.
Toggle SageAttention, FFmpeg, and optional downloads without leaving the page.
Inspect the live JSON plan before starting the install.
Use Extra Settings for in-memory overrides without mutating the embedded defaults.
What it does
You run the standalone installer binary. It opens a local web UI, asks you a handful of questions upfront, then handles everything else unattended:
Downloads and configures a fully portable Python 3.12 environment isolated inside your ComfyUI folder
Clones ComfyUI at the latest stable release tag or a specific version you choose
Detects your GPU and installs the exact right PyTorch build for your hardware
Optionally installs SageAttention, FFmpeg, and a curated set of custom nodes
Creates a ready-to-launch
run_comfyuistarter in your ComfyUI folderOn subsequent runs, detects what's already installed and asks whether to update, refresh, or leave it alone
Quick Install
The release binaries are the simplest path for most users. Download the file for your OS, put it in the folder where you want the installer state to live, and run it directly:
Linux:
./dasiwa-installer-linux-amd64Windows:
dasiwa-installer-windows-amd64.exe
./dasiwa-installer-linux-amd64On Windows, double-click dasiwa-installer-windows-amd64.exe.
If you want a direct download from the latest GitHub release:
curl -L -o dasiwa-installer-linux-amd64 \
https://github.com/darksidewalker/dasiwa-comfyui-installer/releases/latest/download/dasiwa-installer-linux-amd64$u='https://github.com/darksidewalker/dasiwa-comfyui-installer/releases/latest/download/dasiwa-installer-windows-amd64.exe'; $o='dasiwa-installer-windows-amd64.exe'; if (Get-Command curl.exe -ErrorAction SilentlyContinue) { curl.exe -fL --retry 5 --retry-delay 2 -o $o $u } else { Start-BitsTransfer -Source $u -Destination $o }The app opens a local browser page and runs the native Go install engine. The UI, default config, placeholder assets, README, and license are embedded in the binary, so users do not need Python scripts, shell scripts, PowerShell scripts, config.json, a node-list text file, or a cloned copy of this repository next to the executable. On first run it creates a local .dasiwa/ bootstrap directory for uv, the managed Python runtime, and cache files, then keeps all ComfyUI packages inside ComfyUI/venv/.
▶️ After installation
Open the ComfyUI folder that was created and launch:
Windows — double-click
run_comfyui.batLinux — run
./run_comfyui.sh
Your browser will open automatically. The first launch may take a minute while ComfyUI pulls its frontend assets.
❌ Do not run as Administrator — it is not needed and can cause permission issues.
🛠️ Requirements
GPU: NVIDIA (GTX 10-series or newer), AMD (RX 6000+) or Intel Arc
Disk space: ~20 GB free for a full install including models
Internet: Required during installation
Python / Admin rights: Not needed — everything is self-contained
Git is installed automatically on Windows if missing.
⚖️ Disclaimer
Free community tool, provided as-is. The author is not responsible for system errors, data loss, or hardware issues. AI workloads are demanding — make sure your cooling is adequate. Third-party software downloaded by this installer (ComfyUI, PyTorch, etc.) is subject to its own respective licenses.

