gitmyhub

ai_game_workbench

TypeScript ★ 62 updated 10d ago

Local-first AI game asset workbench for 2D character workflows

A local web app for creating 2D game character animations and pixel-art sprite sheets using AI image and video generation APIs you bring yourself, with direct export to the Godot game engine.

TypeScriptNode.jsCloudflaresetup: moderatecomplexity 3/5

AI Game Workbench is a local desktop tool for creating 2D character animation assets for games using AI image and video generation models. The README is in Chinese. It runs as a local web application: you start it on your own machine, open it in a browser, connect your own API keys for the AI services you want to use, and work from there. No built-in keys are included, so you bring your own access to image and video generation APIs.

The tool has two main modules. The first handles high-resolution 2D character creation, covering common animation states like walking, idle, running, jumping, and attacking. For each state it can generate still frames, produce video clips, extract individual frames from those videos, remove green-screen backgrounds, handle looping, and preview the result. It also exports assets in a format compatible with the Godot game engine. The second module handles pixel-art characters at a fixed 64x128 pixel size, with similar steps: generate a base pose, produce a walk cycle, handle chroma keying, slice the sprite sheet, and output individual transparent frames.

For video generation to work, the tool needs images accessible via a public HTTPS address. To handle this without requiring a server, the launcher automatically creates a temporary Cloudflare tunnel that exposes your local files to the internet under a randomly assigned URL. The tunnel is only active while the tool is running.

On Windows, a prebuilt launcher executable handles startup: it starts the local API server, starts the web frontend, creates the tunnel, and opens the browser. Developers can also run from source using Node.js and npm. API keys are stored in the browser's local storage and never written to the source code or configuration files. All generated images, videos, and exports are saved in a local storage folder on your machine.

Where it fits