Recommended path
ComfyUI Desktop
Use this for the normal Windows app install. It opens like a regular app and includes ComfyUI Manager by default.
Download ComfyUI DesktopComfyUI architecture LoRA setup
ComfyUI Architecture LoRA Setup.
Set up the local ComfyUI stack used by ALTERA2ION packages: Windows install, model files, LoRA placement, workflow JSON, and the package installer.
Local workflow stack
Built for local generation.
A Windows workstation with an NVIDIA GPU is recommended. Once ComfyUI and the base models are ready, the installer places the encrypted LoRA and loader in the expected folders.
GPU
NVIDIA recommended
VRAM
8GB minimum
RAM
32GB minimum
OS
Windows 10/11
Windows local workstation
NVIDIA recommended
Three model downloads
Installer included
After setup
Choose the workflow.
One workstation can run the interior, exterior, and form studies once the model files and loader are in place.
01 / Install ComfyUI
Install ComfyUI.
Choose Desktop for the normal Windows app. Portable is for users who prefer an unzip-and-run folder.
Folder-based install
Portable NVIDIA
Use this if you prefer downloading a folder, unzipping it, and running ComfyUI from inside that folder. For NVIDIA Windows machines.
Not sure? Choose ComfyUI Desktop.
It is the simplest route and includes ComfyUI Manager.
02 / Download models
Download the model files.
Keep each file name unchanged and place it in the exact ComfyUI folder shown.
01
Flux 2 Klein 9B FP8
Main diffusion model. Download this first.
Download Flux 2 Klein 9B FP8Place this file in
ComfyUI\models\diffusion_models\flux-2-klein-9b-fp8.safetensors
02
Qwen text encoder
Text encoder file used by the workflow.
Download Qwen 8B FP8 mixedPlace this file in
ComfyUI\models\text_encoders\qwen_3_8b_fp8mixed.safetensors
03
Flux 2 VAE
VAE file required by the calibrated graph.
Download Flux 2 VAEPlace this file in
ComfyUI\models\vae\flux2-vae.safetensors
Qwen note
Use the direct Qwen button above. Do not download shard, chunk, or pointer files from the repository page, and do not manually merge files.
Keep the tested stack
Do not rename or replace these model files. 16GB+ VRAM is recommended for heavier scenes, while 8GB remains the launch minimum.
03 / Install package
Run the package installer.
After ComfyUI and the model files are ready, run the installer inside your purchased package.
01
Open 01_INSTALLER
Inside your downloaded package, open the folder named `01_INSTALLER`.
02
Run package installer
Windows: double-click `run_package_installer.bat`. macOS: open `run_package_installer_mac.command`. The installer copies the `.a2enc` file into `ComfyUI/models/loras` and installs the required custom-node packages.
03
Open the workflow
In ComfyUI, load `01_WORKFLOW.json` from the package workflow folder. Queue once, then approve the ALTERA2ION LoRA Loader in the browser with the same account that owns the package.
If the ALTERA2ION node is red
Check the LoRA file location.
The encrypted LoRA must sit directly inside `ComfyUI/models/loras`, not inside a nested `05_LORA folder`.
After fixing the location, restart ComfyUI and queue again. If the browser does not open, rerun the installer.
Task-specific workflow package
Curated by ALTERA2ION for the task.
Each package combines a calibrated workflow, encrypted fine-tuned LoRA, installer, guide, and example structure for a specific architectural use case.