Block reference parts of images just like text with Obsidian Excalidraw

Linking a zoomed-in “piece” of an image to the exact place it came from has long been a pain in Obsidian-style note systems: copying a big image and then manually cropping parts breaks the connection, forcing readers (and authors) to hunt for provenance. Obsidian Excalidraw’s new block/section reference transclusion feature fixes that by letting users name a region inside a drawing and then embed only the relevant portion back into markdown—while keeping clickable navigation back to the source area in the full image.

The workflow starts with Excalidraw’s embedding options for markdown. Excalidraw supports three embedding modes: a markdown text embed, a text-area embed (the referenced text element plus surrounding padding), and an element-group embed (the entire group containing the referenced element, with nothing else from the drawing). On top of that, the system distinguishes between block references and section references, and it offers two drawing export formats: SVG (default) and PNG. Several of the more precise behaviors—especially around “area” transclusion—depend on using SVG.

In practice, a control-click on a referenced text element creates a link that can be pasted into markdown. Adding parameters like “area” or “group” changes what gets transcluded when the link is rendered. A block reference shows only the referenced text element itself. An area reference transcludes a cut-out around that element, including padding so the context isn’t lost. A group reference transcludes the entire group the element belongs to, which is useful when the “piece” is meaningful only as part of a larger composite (for example, a labeled figure assembled from multiple shapes). The transcript also highlights a subtle but important behavior: group transclusion is limited to elements actually in the group. If a shape sits visually inside the larger drawing but is not part of the referenced group, it won’t appear when transcluding the group.

The link-copying UI makes these reference types fast to generate using modifier keys: control/command-click copies a group reference, shift-click copies an area reference, and a plain click copies a block reference. The transcript notes a surprising edge case: when only a block reference is copied, transclusion may include less than expected because of how Excalidraw structures markdown paragraphs—sometimes a “block” is only the final paragraph containing the reference, not the entire section above it. Switching from SVG to PNG reveals another limitation: with PNG embeds, area transclusion breaks down. Instead of respecting the area cut-out, the system can end up transcluding the entire image, and the “area” option is effectively ignored.

Finally, the transcript warns about reference validity. Copying a block reference for a non-text element can produce errors if that block reference can’t be found in markdown. Prepending “group” can avoid the failure by transcluding a valid group context, though it changes what appears (e.g., including padding around the target region).

To make a complex annotated image usable—using a Peter Bruegel painting example—the creator applies practical canvas tricks: resizing the base image to fit sticky-note label boxes, locking the image so annotations don’t shift it, and setting annotation strokes to transparent after finishing so they can be toggled back to visible quickly with a select-all and color change. The result is a more reliable, navigable way to connect image regions to notes without losing the “big picture” context.