Memory Garden

Original Memory Garden concept and Processing sketch by Alexander Bie. Web adaptation built with Claude Code.

Turn personal memories into AI-generated flower art — all in the browser.

Each memory is analyzed for five emotional dimensions (valence, impact, togetherness, motion, length), which drive the shape, color, and texture of a procedurally generated flower. The flower is then transformed into a stylized artwork using an AI image model.

Quick Start

1. Install uv

Mac:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows (PowerShell):

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Set up API key

cp .env.example .env

Edit .env and replace your_replicate_api_key_here with your Replicate API token.

3. Run

uv run --with flask --with replicate server.py

Your browser will open http://localhost:8080 automatically.

How It Works

Your Memories (text)
      |
      v
DeepSeek V3.1 (LLM)          -- analyzes emotional features
      |
      v
5 Emotional Dimensions         -- valence, impact, togetherness, motion, length
      |
      v
p5.js (browser)                -- generates procedural origami flower
      |
      v
Nano Banana Pro (diffusion)    -- transforms into stylized artwork
      |
      v
Final Image

Project Structure

flower/
  server.py                    Flask backend + Replicate API proxy
  static/
    index.html                 Web UI
    sketch.js                  p5.js flower renderer (ported from Processing)
  garden.json                  Memory data (emotional features)
  .env.example                 API key template (copy to .env)
  MemoryGarden/
    MemoryGarden.pde           Original Processing sketch
  TUTORIAL.md                  Student tutorial

Documentation

• TUTORIAL.md — Step-by-step student guide