AI CREATION CANVAS · BUILT ON COMFYUI
An open-source canvas for AI images and video. Bring models, media and prompts together. Create with ComfyUI, then take the result somewhere new.
TakeBoard
A reference becomes an image. An image becomes a shot. Build on what you make, or explore a different direction alongside it.
Add your references and write a prompt.
Choose a workflow and generate on the canvas.
Connect a result to your next creative step.
Add a recommended workflow or bring your own from ComfyUI. Choose a compatible model for the image or video you want to make.
Models and nodes are installed separately. Available inputs depend on the workflow.
Work with prompts, references and generation controls in one place. Explore variations without leaving your project.
Run ComfyUI on your computer or connect to a remote machine. The canvas stays with you; you choose where generation happens.
42 seconds. From a starting frame to a result—and the next scene.
Actual product capture · Generation wait omitted
About this demoThe desktop app, for your computer.
Requires your own ComfyUI and models for generation. The app is currently primarily in Chinese. Beta packages are not Apple-notarized or commercially Windows-signed.
Installation and opening help →Connect it in Settings, add a compatible workflow and create your first shot.
Create a project and bring in an image. Explore the canvas before connecting a generation service.
TakeBoard brings image and video creation onto a connected canvas, using your own ComfyUI to run the workflows. Use ComfyUI's editor when you need to change the underlying node graph. TakeBoard is an independent project, not affiliated with Comfy Org.
Start with a recommended template or import your own Workflow JSON, API Prompt JSON or PNG with workflow metadata. TakeBoard checks dependencies and inputs. Custom workflows may need parameter mapping and trust confirmation; unsupported graphs can be exported as API format from ComfyUI and imported again.
Local project management does not require registration. TakeBoard supplies no hosted GPU service. Your models, remote infrastructure and any third-party services have their own requirements and costs.
A local project saves to your chosen folder. Remote ComfyUI receives the required inputs, and TakeBoard collects outputs into that project. Opening a remote TakeBoard project is a separate option: its files stay on the remote host.
Recommended adaptations include Qwen Image, MiniMax H3, Wan 2.2 and LTX 2.3. You can also import custom workflows. Available inputs and generation options depend on the workflow and the models and nodes installed on your device. TakeBoard is currently in beta.
Current packages are not Apple-notarized or commercially Windows-code-signed. Read the installation guidance; do not globally disable system protections to try an app.
Product descriptions, screenshots and an interaction video for your next article, tutorial or review.
