What is open-canvas-tv?

Open Canvas is a browser based tool for building AI image and video creation workflows on an infinite visual canvas instead of running one prompt at a time. You place nodes on the canvas, connect them together, and each node represents a step such as uploading a source image, staging a 3D scene, generating an image, or generating a video from that image. The output of one node can feed directly into the next, so a whole multi step creative process stays visible and reusable instead of disappearing after a single generation. A notable piece is the built in 3D Director, which lets you stage characters, poses, camera angles, and simple shapes like cubes and spheres before spending any credits on actual generation. The screenshot from that staging area can then be fed into an image or video generation node as a reference. Beyond that, the canvas supports multiple image models and multiple video models through one consistent interface, covering things like text to image, image to video, and first frame to last frame video generation, each with its own supported resolutions and clip lengths. The project describes itself as local first: your API key and canvas projects are stored in your own browser using local storage and IndexedDB rather than on a separate application server or account system. Actual image and video generation happens remotely through the SJolt API, and the generated media is delivered from SJolt's content delivery network, so local first here refers to where your project data and credentials live, not where the AI models run. Open Canvas itself is open source under the MIT license and free to run, but the underlying model calls are billed to your connected SJolt account, so using it in practice costs money based on how much you generate. To run it locally you need Node.js 20 or newer and pnpm, after which you clone the repository, install dependencies, and start the development server, then connect your own API key in the browser.