Synese-node image

Harbor Bai

Selected Works

About

Synapse

Node-based image generation workflow

Role:
Web developer
UX designer

When:
Aug 2025
1 week solo project

Focus:
AI/human collaboration
UI/UX

Tools:
Replit
Gemini API
HTML
CSS

Overview:

Most AI image tools are trapped in LLM-style chat UIs. Synapse replaces scrolling prompts with a visual graph of nodes, references, and parameters—so users can see, branch, and iterate like they do in Figma or Houdini.
What would an AI image tool look like if it were designed for designers, not chat?

Painpoint:

Today’s AI image tools inherit the chatbox(google gemini, chatGPT etc): long prompt threads, hidden states, weak reproducibility, and limited branching. Designers need:

Spatial awareness of their pipeline (not a wall of text)
Determinism & versioning to revisit exact results
Parallel exploration (branching, forking, A/Bs)
Reference-centric workflows (images, masks, styles as first-class objects)

The Concept — “Canvas” : A node-based studio where every element is visible and editable:

Why it matters?

Faster learning loop meets creative branching and design rigor—where you see and edit the pipeline itself, explore multiple visible paths, and turn prompts into structured, reusable systems.

What’s New vs. Chat UI:

Shift from linear to spatial, opaque to explicit, and single-track to multi-branch—where exploration happens visually, parameters and references stay transparent, and ideas diverge widely before converging with clarity.