Image Relationship Graph: How to See How References Connect

By refern | Last updated: June 2026

The short answer: refern's graph view maps your entire visual library as a navigable network. Every image, folder, canvas, group, and tag becomes a node. Typed links between them become edges. Click any node to jump directly to that file or folder. Use search operators like derived:true or linked:true to filter by relationship type without opening the graph at all.

Most reference managers treat your library as a flat pile of files with folders on top. That works fine at a few hundred images. At a few thousand, patterns emerge that a flat list cannot show: this character sheet was traced from that reference photo, these mood boards pull from the same two folders, this canvas reuses images that also appear in three other projects. A relationship graph makes those patterns visible.

Before you start

What you need:

• refern installed (30-day free trial, then $30 one-time at refern.app, going to $35 about two months after launch)
• At least one workspace open with some images indexed
• Optional: a canvas file with images on it, which gives you placed-in-canvas links to explore

If you are new to refern, the glossary entry on what a reference manager is covers the basics before you dive in.

Step 1: Open the graph view

Click the graph icon in the left sidebar (or press the keyboard shortcut shown in the sidebar tooltip). The graph opens as a full-screen overlay.

On first load, refern builds the node and edge layout using a force-directed simulation. Nodes spread out until they reach a stable arrangement. Large libraries with many links may take a moment to settle.

What you see:

• Each image, folder, canvas, group (link container), and tag appears as a circle (node)
• Lines between nodes are the typed links that exist in your library
• Nodes cluster naturally around their most-connected neighbors

The graph is live. If you add a link or create a new canvas while the graph is open, the change appears without reloading.

Step 2: Understand the four link kinds

refern tracks four typed relationships. Each kind appears as a different visual treatment in the graph so you can distinguish them at a glance.

member-of: an image belongs to a group (a fan-card cluster in the grid). One image can belong to only one group. The group node appears between its members and its parent folder.

derived-from: an image was cropped or saved as a new file from a source image. When you use refern's crop tool and choose "Save as new," the new file automatically gets a derived-from link pointing back to the original. You can trace the entire crop lineage forward and backward in the graph.

placed-in-canvas: an image appears on a canvas file. Every time you drop an image onto a canvas and save, refern writes a placed-in-canvas link from that image to the canvas. An image that appears on five canvases has five placed-in-canvas edges.

cross-reference: two images are manually linked as related. You create these yourself from the sidebar (or with the keyboard shortcut) when you notice that two references are conceptually connected but have no automatic provenance relationship. Cross-references are pairwise: linking A to B also links B to A.

Understanding these four kinds matters because you can filter on them. More on that in Step 4.

Step 3: Navigate the graph

Click a node to select it. The right sidebar updates to show that entity's metadata (name, tags, ratings, linked references). If the entity is an image, a thumbnail preview appears.

Double-click a node to navigate to it. The graph closes and refern opens that file or folder in the main view. This is the fastest way to jump from a discovered connection to the actual asset.

Scroll to zoom in and out. Pan by dragging the background. The graph has no hard boundary; nodes will be wherever the simulation placed them.

Use the degree filter to focus on a neighborhood. When you select a node, you can dial the visibility down to show only first-degree connections (directly linked), second-degree (one hop away), or third-degree and beyond. This is useful in a large library where the full graph is too dense to read. Narrow to the focal node and its immediate neighbors.

Hover a node to see its name and link count without selecting it.

Step 4: Use search operators to filter by relationship (without opening the graph)

The graph is the visual tool. For precise querying, the search bar is faster.

Find everything derived from a source:

derived:true

Returns every image in your library that has a derived-from link. These are all the crops and save-as-new files you have created.

Find images that appear on at least one canvas:

placed-in-canvas:true

Or use the linked operator to find images with any link:

linked:true

Find images with no links at all:

linked:false

Useful for auditing: these images exist in your library but have no relationship to anything else. They may be orphaned references you forgot about, or genuinely standalone assets.

Find everything linked to a specific image:

linked-to:<entity-id>

The entity ID appears in the URL when you open a file, or in the metadata sidebar. This operator returns every image that is directly connected to that specific file by any link kind.

Combine operators with other filters:

derived:true tag:anatomy rating:>=4

Returns only highly-rated anatomy references that are crops or derived files. Typed search and typed links compose cleanly.

Step 5: Build links intentionally

The graph is most useful when you build links with a purpose, not just as a side effect of cropping and canvas work.

Use cross-references to connect references you plan to study together. If you have a lighting reference and a color palette reference that always inform the same kind of work, link them. When you open one in the sidebar, the "Linked to" section shows the other immediately.

Use the crop tool to capture details with provenance. Instead of saving a crop to a separate folder and losing track of its origin, use refern's crop tool with "Save as new." The derived-from link is written automatically. Six months later, when you find the crop and wonder where it came from, the graph shows you instantly.

Use canvases as project organizers. An image that appears on a canvas for Project A and a canvas for Project B has two placed-in-canvas links. The graph shows both canvases connected to that image. You can see at a glance which images are shared across projects and which are used only once.

Use groups for compositional clusters. When you group images in the grid (fan-card mode), the group node appears in the graph between its members and their folder. Grouping is not just a visual grid feature; it becomes a node in your relationship network.

Step 6: Read the Linked References sidebar

You do not need to open the full graph to see an individual image's connections. The sidebar's "Linked References" section shows three sections for the selected image:

Cropped from: the source image, if this image was derived from one. One click opens the original.

Placed in canvases: every canvas this image appears on. One click navigates to that canvas with the image selected.

Linked to: every image manually cross-referenced to this one. These are the relationships you built intentionally.

The sidebar is the low-friction everyday view. The full graph is for discovery and overview.

How refern's approach compares to other tools

It helps to understand what problem this feature is solving and why other tools in this space have not solved it the same way.

Obsidian has the most well-known relationship graph in personal software, but it is built for text notes and wikilinks. Obsidian's graph is genuinely excellent at what it does: mapping how your written ideas connect to each other. It does not apply to images. Obsidian has no native image tagging, no thumbnail gallery, no color search, and no image-to-image provenance. Artists who use Obsidian for text notes and want similar graph-based navigation for their visual references are exactly the audience refern was built for. The two tools are complementary; refern is the visual side of the same philosophy. (Obsidian's core app is free; their Sync add-on costs $4 to $5 per month as of 2026.)

Eagle ($34.95 one-time as of 2026) is the most capable image organizer in this category, with 99 to 108 native format previews, font management, and a mature plugin ecosystem. Eagle has no graph view and no entity linking. There is no way in Eagle to see which images were cropped from which source, which images appear on a canvas, or how files and folders relate to each other beyond the folder hierarchy. Eagle is a strong choice if format breadth and plugin extensibility matter more than relational navigation. If the relationship layer is important to your workflow, Eagle does not have it.

Common problems and fixes

"The graph is too dense to read." Select a node that interests you and reduce the degree filter to 1 or 2. This shows only immediate neighbors and makes the local neighborhood readable. For very large libraries, treat the graph as a neighborhood explorer rather than a global map.

"I do not see any links in my graph." Links are created in three ways: automatically when you use the crop tool with "Save as new" (derived-from), automatically when you save a canvas with images on it (placed-in-canvas), or manually via the Link button in the sidebar (cross-reference) and the Group action in the grid (member-of). If you have never done any of those, the graph will show nodes with no edges. Start by adding a few images to a canvas, saving it, and reopening the graph.

"I cropped an image but there is no derived-from link." The derived-from link is only written when you use refern's built-in crop tool with "Save as new." If you cropped in an external application and imported the result separately, refern has no way to know the provenance. You can create a cross-reference link manually to represent the relationship.

"I added an image to a canvas but the placed-in-canvas link does not appear." The canvas must be saved after adding the image. Links are written on save. Open the canvas, add or confirm the image is there, and save.

Next steps

Once you are comfortable reading and building your relationship graph, these pages cover related features in depth:

• refern vs Eagle: 2026 comparison covers the full feature difference between the two most common asset managers in this category
• refern vs PureRef alternatives covers the canvas and overlay use case if you are coming from a reference-pinning workflow
• What is a reference manager if you want foundational context on how these tools fit into a creative workflow

The relationship graph is the feature that separates a static image pile from a navigable knowledge base for your visual references. Building links consistently over time, even just the automatic ones from cropping and canvas work, compounds into a map of your creative process that is genuinely useful to look at months later.