RTX 5060 Ti 16GB ComfyUI Setup

ComfyUI is the node-based image-gen UI that power users gravitate toward. Install on the RTX 5060 Ti 16GB at our hosting:

Contents

• Install
• Model folders
• Run server
• Remote access
• Custom nodes

Install

git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
uv venv --python 3.12
source .venv/bin/activate
uv pip install torch --index-url https://download.pytorch.org/whl/cu126
uv pip install -r requirements.txt

Model Folders

Drop weights into:

• models/checkpoints/ – SDXL, SD 1.5, FLUX base
• models/loras/ – LoRAs
• models/controlnet/ – ControlNet weights
• models/vae/ – custom VAEs
• models/clip/ – FLUX text encoders (T5, CLIP-L)
• models/unet/ – FLUX UNet (fp8 variants)

Run

python main.py --listen 0.0.0.0 --port 8188

Default flags work well on Blackwell. For FP8 FLUX weights add --fp8_e4m3fn-unet.

Remote Access

Two options:

• SSH tunnel: ssh -L 8188:localhost:8188 user@server then browse http://localhost:8188 locally. Safe default.
• TLS proxy: Caddy or nginx with basic auth, gives public URL. Needed if team members can’t SSH.

Custom Nodes

Install ComfyUI-Manager for the best experience:

cd custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Manager
# Restart ComfyUI

Now you can install nodes (IP-Adapter, ControlNet, upscalers) from the in-UI manager.

Recommended starter node packs:

• ComfyUI-Impact-Pack (detailer, face enhance)
• ComfyUI-IPAdapter_plus
• ComfyUI_essentials (utility nodes)
• ComfyUI-AnimateDiff-Evolved if doing animation

See also: SD setup, A1111 setup, FLUX benchmark, image studio.