Apps / Productivity
Stable Diffusion
Images, generated here
The AUTOMATIC1111 web interface for Stable Diffusion — text to image, image to image, inpainting, upscaling — running on your own hardware. No subscription, no queue, no content filter deciding what you may ask for, and nothing you generate is uploaded anywhere.
CategoryProductivity
Memory~8GB
Module
stablediffusionInstalling it
Open the dashboard, go to Apps, press Add on it and apply the change. Or from a terminal on the box:
sudo homebox install stablediffusionWhere to open it
| Service | Port | First login |
|---|---|---|
| Stable Diffusion | :7860 | No account by default. Anything that reaches the page can generate on your hardware. |
Open it at your box's address with that port — the dashboard links to it directly once it is running.
Good to know
- This wants an NVIDIA GPU. On a CPU a single image takes several minutes instead of several seconds, which is the difference between a tool and a demonstration
- A GPU needs the NVIDIA Container Toolkit on the host and a deploy.resources.reservations.devices block here — deliberately not added, because on a box without one it stops the container from starting at all
- Models are 2-7GB each and are not included. Put .safetensors files in the module's models directory and restart
- The first start downloads several gigabytes of Python dependencies before the page appears. Give it time before deciding it is broken