Drawing with AI

The Vellum canvas has an AI panel. Describe what you want and it draws it; select shapes and ask for a change and it edits them.

The Vellum AI panel: Draw, Convert and Explain tabs above a prompt box, with example questions and the credit balance at the foot

Draw, Convert, Explain

The panel opens with three modes. Draw produces shapes. Convert turns an image of an existing diagram into an editable one, opening the import dialog rather than starting a conversation. Explain answers questions about the diagram in front of you, such as "what does this show?" or "where's the bottleneck?", and will not touch the drawing unless you ask it to.

The panel header names the current scope, so you can see whether you are asking about one selected shape or the whole canvas before you send.

Generating

Describe the system in prose. Be specific about the parts and how they connect: the model draws what you name, and invents structure where you leave gaps.

Generation produces an ordinary diagram. Every shape is editable, and there is no separate "AI layer" to unpick later.

Editing

Select shapes and describe the change. With a selection, the edit is confined to it; with nothing selected, the whole diagram is in scope.

Expect to finish it yourself

Generation gets you a workable draft, not a finished diagram. Layout, grouping and colour still need your judgement.

Cost

Generation and editing draw on your allowance or credits. A failed generation still costs, because the work was done even if you did not like the result.