Type @ in any prompt input to inject a canvas element as context. The app shows a picker; pick an image, video, note, or group and it becomes part of the request.
Where it works
- Generate Image prompt
- Generate Video prompt
- Agent Dock chat
How it works
- Type @ anywhere in the textbox.
- A floating picker shows the currently selected elements first, then every element on the board.
- Click the element you want, or type to filter by name.
- The element appears as a chip in the prompt (for example
@hero-render.png). Multiple mentions are stacked.
When you submit, the app:
- For images and videos — attaches them as reference media to the generation request.
- For notes — pastes the note’s text content into the prompt.
- For frames or groups — attaches every child element.
Why use it
- No re-upload — references are already on the board.
- Iterate fast — generate, mention the result, iterate without dragging.
- Combine sources — mention a photo and a note containing brand colours in the same prompt.
Mention vs reference slot
The Generate Image panel has a dedicated Input Image slot. For image-to-image with a single reference, that slot is the simpler path. Use @-mentions when:
- You want multiple references.
- You also want to inject a note’s text.
- You are working in the agent dock where there is no separate slot.
Limits
- The picker shows the active board only.
- Very large videos (multi-GB) may take a few seconds to read for the agent context.
- A maximum of about 10 mentioned items per prompt — beyond that, the LLM’s context window starts to crowd out the prompt itself.
Leave a Reply