AI Art

paused

CLI tool for local AI art generation using Stable Diffusion 2.1, optimized for Apple Silicon M3 GPU with Metal Performance Shaders

Tech Stack

PythonPyTorchDiffusersStable Diffusion 2.1Apple Metal

TL;DR

Text prompt → Stable Diffusion → Pretty picture

generate.py
$ python generate.py \
"a lone astronaut standing
on the edge of a massive
crater on Mars, Earth visible
in the pink sky, cinematic
lighting, hyper detailed" \
--steps 50 --guidance 9
Loading SD 2.1...
Using MPS (Apple Silicon)
Generating (50 steps)...
✓ Saved to output/

CLI prompt input

AI generated art sample

Generated output (512×512)

What We (✨Claude✨) Built — an honest confession

Phase 1

The Real Reason

This project exists because I needed something to test the /add-work and /log-work slash commands I built for this site.

I spent way more time perfecting the work tracking system than actually generating art. Turns out I'm not good at art — even when the AI does most of the work.

Time ratio: ~10% writing the CLI, ~90% testing slash commands with it as the guinea pig

Phase 2

But It Actually Works

The CLI does run Stable Diffusion 2.1 locally on Apple Silicon. No cloud APIs, no subscriptions — just your M-series GPU doing the heavy lifting via Metal Performance Shaders.

It's nothing fancy: pass a prompt, wait ~30 seconds, get a 512×512 image. But there's something satisfying about running AI models locally instead of paying per generation.

Stack: Python, PyTorch, Hugging Face Diffusers, ~4GB VRAM on M3