Infinite canvas for neural networks — how AI CanvasPro turns working with AI into visual creativity

Imagine you're working on a complex project: you need to generate a character, write a script for them, voice the lines, and turn it all into a short video. Usually this turns into a nightmare of browser tabs: ChatGPT here, Midjourney there, ElevenLabs in a third window, and editing in a fourth app altogether. A Chinese developer under the nickname ashuoAI decided that enough is enough and built AI CanvasPro.

This is an Electron desktop application that turns the generative process into working with nodes on an infinite canvas. If you've ever opened Blender (Geometry Nodes) or DaVinci Resolve (Fusion), you'll immediately understand the logic.

https://raw.githubusercontent.com/ashuoAI/AI-CanvasPro/master/images/favicon.svg

What is it anyway

AI CanvasPro is a visual editor, where nodes are used instead of layers or chats. You place a node for text, connect it to an image generation node, and send the result to a video or upscaler node. The project doesn't try to replace neural networks itself—it works as a "glue" for different APIs and local tools.

The nicest part is that the project supports not only cloud APIs like OpenAI or Gemini, but also deeply integrates with ComfyUI. This means you can pull your local workflows right onto the canvas and combine them with cloud models.

How it works in practice

The main feature here is data connectivity. In a regular chatbot, context is limited to one dialogue. In AI CanvasPro you can use the symbol @ right in the prompt input field to reference the output of any other node on the canvas. For example, take a character description from a text block and pass it to an image generator.

Main node types

You can place a dozen different node types on the canvas. Here are the ones that seemed most useful to me:

• Material source. You can simply drag a file from disk here—an image, sound, or video.
• Generators. Separate blocks for text (GPT-4, Gemini), images (via RunningHub or APImart), and video.
• Browser node. Allows you to pull content directly from the web and send it straight for processing.
• Editing tools. Built-in mini-editors for video trimming, background removal from photos, and even working with 360-degree panoramas.

Audio work is implemented interestingly. There's a separate "audio studio" where you can clone a voice or separate vocals and background music. All of this happens within one project, without endless exports and imports.

Technical side and installation

The app is built on Electron, so it's cross-platform. There are builds for Windows and macOS (including Apple Silicon).

An important note about "openness": the author posted the project on GitHub, but it's not Open Source in the classic sense. The license prohibits commercial use without permission and reselling the app as SaaS. There are no restrictions for personal use and experiments.

To get started, you need to:

1. Download .exe or .dmg from the Releases section.
2. Go to settings (profile icon in the bottom left).
3. Enter your API keys. Both specific Chinese providers (RunningHub, Volcano Engine) and the standard OpenAI API format are supported.

Who is this for

I see two main use cases. The first is content creation for social media or YouTube. When you need to quickly run an idea through the chain "text -> image -> voiceover -> video," a visual canvas saves a lot of time.

The second use case is prototyping complex pipelines. If you're a developer and want to quickly test how a combination of several models will work, throwing nodes on a canvas is much faster than writing a Python script.

AI CanvasPro is a solid tool for those tired of chatbot interfaces. The project has its rough edges—for example, part of the interface and documentation is in Chinese (though the main functions are intuitively understandable), and the license is quite strict. But the "AI operating system" concept on a canvas is implemented very well.

If you actively use ComfyUI or often work with multiple neural networks simultaneously, it's definitely worth trying. At minimum, it's a great way to organize your generations.

You can check out the project and download the latest version in the ashuoAI/AI-CanvasPro repository.