AI & Creative Tools

Arty - Digital Art Assistant

A team-built digital art assistant combining reference search, BLIP-assisted image understanding, Stable Diffusion generation, and artist workflow tools.

Arty was a UQ DECO3801 capstone project for digital illustrators. The product goal was not to replace the artist; it was to reduce the tedious parts of a concept workflow: finding references, organising ideas, and producing fast visual starting points while keeping creative control with the person drawing.

Arty application screens
A product screen overview covering search, upload, and workspace flows.

Features

  • Reference search - search Pinterest-style image sources from text, uploaded images, or a combination of both.
  • Image understanding - use BLIP captions from uploaded images to improve search queries.
  • AI visualisation - generate starting-point imagery through Stable Diffusion and ControlNet.
  • Creative workspace - keep references and generated concepts beside the artist's canvas.
  • Account/history support - persist users, search history, and generated workflow data with PostgreSQL.

My role

I was the Machine Learning Developer on a six-person team. My main focus was the AI backend: integrating Stable Diffusion, wiring ControlNet options, exposing generation endpoints through Flask, and connecting image-captioning/search flows so artists could move from rough visual references to generated concepts.

Stable Diffusion pipeline diagram for Arty
The generation pipeline connected Flask, Stable Diffusion, ControlNet inputs, and generated image storage.

Technical notes

The Flask backend loads BLIP for image captions, wraps Stable Diffusion WebUI/ControlNet through a Python backbone, and supports both text-to-image and inpainting-style upload flows. The search endpoint can combine a typed query with a caption generated from an uploaded image, which made the reference search feel more visual than a normal keyword box.

The project also includes database routes for user accounts and search history, plus a React frontend with pages for home, search, upload, profile, terms, privacy, and about content. For a student capstone, the interesting challenge was not any single model call - it was turning several brittle AI and web pieces into one understandable artist workflow.

The team

NameRole
William MercadoTeam Leader
Brian ZhangMachine Learning Developer
Hongyingzi LuDesign / Front-End
Shan LiuDesign / Front-End
Yang XiaoBack-End Developer
Ryuto HisamotoBack-End Developer