Windpaint

Introduction

Windpaint is one API for generating images and video, with every output kept as an asset in your project.

Windpaint is an API for generating images and video. You ask for a capability, such as image.generate or video.generate, send a prompt and any input media, and get back a job you can poll. When the job finishes, its outputs are stored as assets in your project, ready to download or to feed into the next job. You pay per output in prepaid credits.

You don’t need to know which model runs a capability. Each capability has a default model, and you can name a specific one when you care. See Concepts for how capabilities, models, jobs, and assets fit together.

Ways to use it

Everything below drives the same API with the same account, credits, and projects. Anything you make in one shows up in the others.

InterfaceUse it for
REST APICalling Windpaint from your own code. Base URL https://api.windpaint.ai/v1.
DashboardGenerating by hand in Studio, browsing assets and runs, managing keys, team, and billing.
windpaint CLIGenerating from a terminal or a script. Uploads local files, waits for jobs, and downloads outputs in one command.
MCP server and agent skillsLetting Claude Code, Codex, Cursor, or any MCP client generate media for you.

What you can build

  • Images on demand in your app. Generate product shots, illustrations, or thumbnails from a prompt and serve them from your own storage. See Images.
  • Short video from a still. Animate an image you generated or uploaded into a 5 second clip. See Video.
  • Multi-step pipelines in one call. A product such as Text to Clip runs several generation steps for you and returns every output.
  • Media from an agent. Give a coding agent a Windpaint key and it can generate, chain, and download assets while it works. See Agents.

Where to start

Windpaint is in early access. New accounts join a waitlist and are activated by hand. Image generation and video generation are live today. The other capabilities you’ll see in the API (image editing, segmentation, upscaling, speech, lipsync, and text) are defined but have no model yet, and return an error if you submit them.