Create
Three surfaces for making an image or a video on purpose — and how to pick between them and a conversation.
Chat can generate media too, and often that's the right way in — you describe what you're after, Endless picks the model and makes it. The Create surfaces are for the other half of the job: when you already know what you want, which model should make it, and how big it should be.
The three ways in
Image
A prompt, optional reference pictures, and every setting the chosen model offers. One run, one set of results, on one page.
Video
The same composer, tuned for clips — and the expensive end of the catalog, so the estimate matters more.
Canvas
An infinite board you share with your workspace. Generate straight onto it, next to what you're already looking at.
Which one to use
Use chat when you're still deciding
You want ideas, variations on a brief, or the picture to come out of a longer piece of work. Endless chooses the model and can keep going after it.
Use Image or Video when you know the shot
You want a particular model, a particular size, a particular reference image. Everything the model can be told is a control, not a sentence you have to phrase well.
Use Canvas when placement matters
You're arranging things — a moodboard, frames of a sequence, variations side by side — and want to generate into that arrangement with other people watching.
They aren't separate libraries. Whatever you make on any of them lands in the same place, and you can start a picture on Image and drag it onto a canvas later.
Before you start: three things that can change what you see
A workspace can switch these surfaces off. Image generation, video generation and Canvas are each separately controllable, and a workspace that turned one off doesn't show it in the sidebar — a saved link lands you back on the workspace home instead. The block holds when a run is requested too, not just in the interface: "Image generation isn't enabled for this workspace". If a colleague has a tool you don't, ask whoever runs the workspace.
Making something is a permission. Every role can look; creating content is a capability a role either has or doesn't. On a canvas that's visible — a role without it opens the board with a "Read-only" badge and no generate bar. See Roles and permissions.
Generating spends credits and the cost varies enormously by model. The composer shows what the run you've set up will cost before you send it. A single video can cost many times what an image costs. See Credits and spending and Models.