← All projects02 / 13
Art-Rium: my own platform for AI images and video
A mobile front end for my ComfyUI workflows. Generate, review, refine and publish, from the phone.
- Year
- 2026
- Context
- Own project
- My role
- Concept, design, full-stack development, workflow design
- Tools
- Python, FastAPI, PostgreSQL, SQLAlchemy, JavaScript, PWA, ComfyUI, Z-Image Turbo, SeedVR2, SAM 3, Wan 2.2, ACE-Step, Ollama, MiniMax
Idea
ComfyUI is powerful but made for the desk. I wanted to work on images wherever I am, and keep a clear process from first prompt to finished print.
What it does
- Generate: images with Z-Image Turbo, with a prompt enhancer that runs locally on Ollama.
- Iterate: render, let a model critique the result, rewrite the prompt, repeat for N rounds.
- Refine in the gallery: enhance, grain, upscale (SeedVR2), crop, variants. Before/after compare, ratings, series.
- Video: clips from images cut to music, transitions with Wan 2.2, cloud renders with MiniMax on a budget. Depth videos and Blender geometry as structure.
- Music: songs with ACE-Step.
- Publish: schedule Instagram posts and draft WordPress articles.
How it is built
FastAPI with PostgreSQL behind it, a light PWA front end for the phone, ComfyUI as the render engine on my desktop. A listener follows ComfyUI’s progress live, and jobs cut off by a restart are cleaned up on startup.
Art-Rium also feeds other projects: the image banks for Keystroke are rendered here.
The AI-painted tool icons 
Made with Art-Rium 
Made with Art-Rium 
Made with Art-Rium