FalSprite

Game-ready sprite sheet animations from a single text prompt. Built on fal.ai — one API key powers everything.

What it does

Type a character description, pick a grid size, choose animation actions — FalSprite generates a full sprite sheet with transparent background and animated preview.

Pipeline (all on fal.ai):

• nano-banana-2 — image generation (sprite sheet)
• OpenRouter LLM (GPT-4o-mini via fal) — intelligent prompt rewriting with character design + choreography
• BRIA — automatic background removal

Features

• Configurable grid: 2x2, 3x3, 4x4, 5x5, 6x6
• Multi-action selection: pick multiple animation actions per grid row (idle, walk, run, attack, cast, jump, dance, death, dodge, or custom)
• LLM prompt rewrite: your simple prompt becomes a detailed character + choreography direction
• Auto background removal: BRIA removes the background automatically
• Live preview: real-time frame-by-frame animation with FPS control
• Downloads: sprite sheet PNG, transparent PNG, animated GIF
• Reference image: upload a reference to guide the generation
• Showcase examples: curated animated GIF examples

Quick start

git clone https://github.com/lovisdotio/falsprite.git
cd falsprite
npm install
npm run dev

Open http://localhost:8787

You need a fal.ai API key — get one at fal.ai/dashboard/keys. That single key covers all three services (image gen, LLM, bg removal).

Deploy to Vercel

npm i -g vercel
vercel

The project includes Vercel-ready serverless functions in api/ and static assets in public/.

Project structure

├── public/           Static frontend
│   ├── index.html    Main app
│   ├── app.js        Frontend logic
│   ├── styles.css    Styles
│   ├── showcase/     Example GIFs
│   └── select.html   Showcase curation tool
├── api/              Vercel serverless functions
│   ├── generate.mjs  Sprite generation endpoint
│   ├── upload.mjs    Image upload endpoint
│   └── fal/media.mjs Media proxy endpoint
├── lib/fal.mjs       Shared FAL API helpers
├── server.mjs        Local dev server
├── batch-generate.mjs  Batch generation tool (dev)
└── process-showcase.mjs Showcase processor (dev)

Tech stack

• Frontend: Vanilla JS, no framework, no build step
• Backend: Node.js (native HTTP server for dev, Vercel serverless for prod)
• AI: All via fal.ai APIs (nano-banana-2, OpenRouter, BRIA)
• GIF: gif.js (browser) + gifenc (server-side batch)

License

MIT