Explaining Links, Transclusions and Embedding in Obsidian Excalidraw

In Excalidraw inside Obsidian, “linking” is straightforward, but confusion usually starts with the difference between transclusion and embedding—especially because Excalidraw’s behavior often makes them feel interchangeable. The practical takeaway: most objects placed into Excalidraw from Obsidian are not copied into the Excalidraw file; they’re referenced from elsewhere (transcluded). Embedding—saving a full copy inside the Excalidraw file—is largely not supported in Obsidian, and that limitation is intentional.

Linking in Obsidian is simply creating a reference to another resource. Excalidraw supports both text links (typed into a text element) and element links (added to objects like rectangles). Clicking those links navigates to the target—either within Obsidian using Obsidian’s built-in navigation or externally via the device’s default application. There’s also a key precedence rule: if a text element already has a text link, that text link overrides any element-link you try to add to the same text element. For web pages, Excalidraw can optionally resolve a page title using an “iframe” service so pasted URLs can become more readable than raw links.

Transclusion is where the “it looks like it’s there, but it isn’t copied” idea matters. Transclusion means Excalidraw shows content by referencing the original file or a specific part of it. In practice, images placed into Excalidraw are transcluded from the Obsidian Vault, which is why Excalidraw files don’t become self-contained bundles of media. Transclusion can be interactive (called embeddables, such as YouTube videos or web pages) or static (non-interactive images that don’t respond to user input). Excalidraw also supports nesting: a transcluded Excalidraw drawing can itself contain other transcluded drawings, creating multiple reference layers. When upstream content changes, downstream nested drawings can refresh—demonstrated by deconstructing a drawing into separate components (mountains, car, person) and showing how edits propagate through the chain.

Embedding, by contrast, is the “copy everything into the file” approach: saving a copy of an object inside another file so the result is self-contained. That’s the default behavior on Excalidraw.com exports, where exporting to an Excalidraw file format can package images and other assets into the exported output. In Obsidian, embedding is practically not supported for Excalidraw files because Excalidraw files are Markdown-based and large embedded binaries would bloat the Markdown and strain Obsidian’s indexing and performance. From a PKM standpoint, embedding also reduces reusability—separating assets into the Vault keeps them manageable and referenceable over the long term.

The transcript also walks through concrete ways to transclude different asset types: images from the Vault, local files via local URI modifiers (with caveats for multi-device sync and mobile), markdown documents as embeddables or as images, and videos as embeddables or as thumbnails. It ends by reinforcing the core distinction: linking is a pointer, transclusion is a reference-based display (including nested references), and embedding is a full copy that Obsidian largely avoids.