Sequence Generator

A visual node-based tool for building AI image generation pipelines. Chain prompts together, generate batches of images with matrix variations, and iterate on results — all from a drag-and-drop canvas.

What It Does

• Node editor — Connect prompt nodes, modification nodes, and image generation nodes into pipelines
• Image generation — Generates images using Google Gemini models via a Python script
• Prompt modification — Uses OpenRouter (Claude) to rewrite prompts with natural language instructions
• Matrix generation — Define dimension variations and generate a grid of images from a template
• Project save/load — Work is persisted locally in the browser

Node Types

Node	What It Does
Base Prompt	Write a starting text prompt
Modify Prompt	Takes upstream text + your instruction, uses an LLM to rewrite it
Append Prompt	Adds text to the end of an upstream prompt
Image	Generates an image from upstream text (Gemini)
Matrix	Generates a grid of images from dimension variations + a template
Multi Image	Generates multiple images from a single prompt

Requirements

• Node.js 18+
• Python 3 with uv installed
• Google Gemini API key — for image generation
• OpenRouter API key — for prompt modification (optional, only needed if you use Modify Prompt nodes)

Setup

git clone <repo-url>
cd sequence-gen
npm install

Create a .env file in the project root:

GEMINI_API_KEY=your_gemini_api_key_here
OPENROUTER_API_KEY=your_openrouter_api_key_here

Get your keys:

• Gemini: https://aistudio.google.com/apikey
• OpenRouter: https://openrouter.ai/keys

Run

npm run dev

Open http://localhost:3000

How It Works

The app is a Next.js frontend with API routes that call out to generate_image.py for Gemini image generation and OpenRouter for prompt rewriting. Images are saved to public/generated/ and served back to the UI.

Prompt nodes form a DAG — text flows from upstream nodes through edges to downstream nodes, where it becomes the input for image generation or LLM modification.