Tag Reference Images So You Can Actually Find Them (2026)

TL;DR: Build a two-level tag taxonomy, use macros to apply tag bundles fast, and lean on autocomplete so tagging does not slow you down. Done right, you can find any image in a 10,000-image library in under five seconds.

By refern | Last updated: June 2026

The real problem with most reference libraries

You save hundreds of reference images. You remember saving a specific piece of anatomy work, a moody rim-light portrait, or a particular fabric texture. Six months later you cannot find it. You scroll, open folders, give up, and go back to Google.

The problem is almost never storage. It is retrieval. A flat folder of images with no meaningful tags is a write-only archive. You can put things in, but getting them back out requires memory you do not have.

This guide shows a tagging method that scales past a thousand images, stays fast enough to use every day, and works whether you are building your first library or migrating one that already got out of hand.

What you will need: A desktop image manager that supports hierarchical tags. This guide uses refern as the example, since its tag system (hierarchical tags, tag groups, linked tags, macros, and rich-text autocomplete input) is purpose-built for this workflow. The principles apply to any tool with hierarchical tag support.

Before you start: pick your tool honestly

Before building a tag taxonomy, you need a tool that will not become a bottleneck. Here is where the main options stand today.

PureRef is excellent as a per-session overlay while you paint or model, but it has no tagging, no search, and no persistent library. It is not a tool for building a retrievable archive. (Source: pureref.com/handbook/features)

TagStudio has the deepest free tag system in this category: parent-child inheritance, aliases, custom colors, and namespace organization. Searching a parent tag surfaces every child-tagged file automatically. It is free, open source, and genuinely powerful. The honest caveats: it is still in alpha, performance on large libraries can be sluggish (Python runtime), and it has no canvas or color search. If free and open source is a hard requirement for you, TagStudio is worth evaluating seriously.

Allusion pioneered the "hierarchical tags, files stay in place" model for artists and influenced many tools that followed. It was a strong choice when it launched in 2021. A GitHub issue filed in April 2025 titled "Project no longer maintained" confirmed what many users suspected: the project has not shipped a release since February 2023 and 83 open bugs have no maintainer responses. A known memory leak consumes 14.4 GB of RAM while generating thumbnails for 358 images. For a library you plan to use for years, building on an abandoned tool carries real risk.

refern is the option if you want hierarchical tags combined with a canvas (replacing PureRef) and color or visual-similarity search in a single app. It costs $30 one-time (launch pricing, going to $35 about two months after launch), with a 30-day free trial and no account required.

Step 1: Design your tag taxonomy before you start tagging

The single biggest mistake artists make is tagging reactively: you see an image, you type whatever comes to mind, and after 500 images you have 300 idiosyncratic tags that overlap, contradict, and cannot be searched consistently.

Spend 20 minutes designing a skeleton taxonomy first. You do not need to anticipate every tag you will ever use. You need to decide on the top-level categories, because those drive everything below them.

A two-level hierarchy is enough for most libraries

A two-level system (parent tag with children) covers the vast majority of reference library needs without becoming a full-time knowledge management project.

Example skeleton for an illustrator or concept artist:

Subject
  ├── Character
  ├── Environment
  ├── Creature
  └── Object / Prop

Anatomy
  ├── Face
  ├── Hands
  ├── Torso
  ├── Feet
  └── Full Body

Lighting
  ├── Rim Light
  ├── Dramatic / Chiaroscuro
  ├── Soft Diffuse
  └── Backlight

Color
  ├── Warm Palette
  ├── Cool Palette
  ├── Desaturated
  └── High Contrast

Style / Medium
  ├── Oil Painting
  ├── Photography
  ├── Digital Illustration
  └── Sketch / Linework

Mood
  ├── Melancholy
  ├── Heroic
  ├── Playful
  └── Tense

With this structure, searching Anatomy returns all images tagged with Face, Hands, Torso, Feet, or Full Body. Searching Anatomy > Hands narrows to hands only. You get broad sweeps and precise filtering from the same tags.

Rules that make the taxonomy survive contact with a real library

Keep parents broad, children specific. If a parent has only one child, merge them. If a child could be a parent itself (it has natural subdivisions), promote it.

Limit yourself to two levels unless you have a strong reason for three. Three-level hierarchies are harder to apply consistently and rarely improve retrieval in libraries under 50,000 images.

Prefer noun phrases. Tags like "Rim Light" and "Oil Painting" are more stable than tags like "Moody" or "Nice Reference." The latter will mean different things to you in six months.

Do not create a tag until you have at least three images it would apply to. Single-use tags become noise.

Step 2: Create your parent tags in refern

In refern, open any workspace and navigate to the Tags panel. Create your top-level parent tags first. Give each a distinct color to make the hierarchy visually scannable in the tag input.

Child tags are created by setting a parent when you make the tag. You can also drag tags onto each other in the tag panel to establish parent-child relationships after the fact.

Tag groups in refern let you cluster related parent tags visually without making one the child of another. Use tag groups to separate major domains (for example, grouping "Lighting", "Color", and "Mood" tags together as a "Aesthetic" cluster in the sidebar, while keeping them as separate first-level tags in the hierarchy).

Linked tags let you declare a non-hierarchical relationship between two tags. For example, you might link "Rim Light" to "Dark Environment" because they co-occur often in your library. This does not affect search results directly, but it surfaces in the tag panel as a suggestion when you are tagging an image.

Step 3: Use the rich-text tag input and autocomplete

The friction point in any tagging system is the moment of application. If tagging an image takes 30 seconds, you will stop tagging consistently within a week.

refern's tag input is a rich-text field that accepts both free text and inline tags. Type # followed by any characters and the autocomplete dropdown surfaces matching tags ranked by prefix match first, then by how frequently you have used them (rare tags surface lower).

Tab accepts the top autocomplete suggestion without a click. For a tag you use often, this means typing #han then Tab applies "Hands" without lifting your hands from the keyboard.

Tag macros are the high-leverage move. A macro is a named shortcut that inserts a predefined bundle of tags in one action. For example, a macro called "character study" might insert Character, Full Body, Anatomy, and Source: Photo all at once. You define macros in the tag panel. To apply one during tagging, type # and the macro name.

For a library with consistent image types (portrait photography, environment concepts, anatomy references), three to five macros will cover 80 percent of your tagging work. You apply the macro, then add one or two specific tags for what is unique about that image.

Step 4: Tag incoming images at capture time, not retroactively

The best time to tag an image is the moment you save it. At that moment you know exactly why you saved it, what attracted you to it, and what projects it relates to. Two weeks later you have none of that context.

refern's browser extension (Chrome, Firefox, Safari) lets you tag images at the moment of capture. When you hover-save or right-click-save an image from any website, the extension shows a tag-on-save panel. Apply your macro, add one or two specifics, and the image lands in your library already tagged.

For drag-and-drop or paste imports into the app itself, the import staging area shows each incoming image and lets you apply tags before the images are written to disk.

Directory metadata presets are refern's answer to folder-level tagging. If you have a folder called "Anatomy References," you can attach a metadata preset to that folder. Every image moved into or imported into that folder automatically receives the tags defined in the preset. This is the lowest-friction tagging method for large batch imports: import the images into the right folder, and the tags follow.

Step 5: Apply tags retroactively to your existing library

If you have an existing library of untagged images, a complete retroactive tagging pass is rarely the right move. It is exhausting and most artists abandon it halfway, leaving a half-tagged library that is harder to reason about than an untagged one.

A more practical approach:

Tag by batch, not by image. Select all images in a folder that share a common attribute (for example, all images in "Hand References") and apply the shared tags to the whole selection at once using bulk metadata editing. Then go image by image only for the specific distinguishing tags.

Use color search to find batches worth tagging together. refern's color search (hex input or color picker) groups images by dominant color, HSV histogram, and color layout. If you search for warm desaturated tones and 40 images appear, those 40 probably share mood-related tags worth applying together.

Use visual similarity to find clusters. Right-click any image and choose "Find similar" to surface images with similar visual features. Images that cluster visually often share subject or mood tags.

Accept partial coverage. A library where 70 percent of images have good tags and 30 percent are untagged is far more useful than a library where 100 percent of images have mediocre inconsistent tags. Tag your highest-priority images well and let the rest come in over time.

Step 6: Search your tags

Once your tags are in place, here are the primary search patterns in refern:

Tag search: Type tag:hands in the search bar to find all images tagged "Hands." The tag: operator supports exact match or prefix. Searching a parent tag returns all images tagged with any child.

Combined filters: tag:anatomy type:image rating:>=4 finds high-rated anatomy images using inline operators.

Smart folders: A smart folder is a saved search query. Create a smart folder called "Anatomy Reference Pack" with the saved query tag:anatomy and it auto-populates every time you open it. No manual curation. See the glossary entry on what a reference manager is for more on how library-level organization differs from per-project boards.

Full-text search: If you saved a description or source URL alongside the image, refern's FTS5 index searches across all text fields simultaneously. A search for "Sargent portrait" finds any image where those words appear in the name, description, notes, source URL, or tags.

Common problems and fixes

"I have too many tags and none of them feel right." You probably skipped the taxonomy design step and tagged reactively. The fix is not to retag everything. Pick the 10 to 15 most important tags in your library, promote them to top-level parents, and organize everything else under them. Archive or merge the rest.

"My tags overlap. I have 'Portrait', 'Face', and 'Character' all meaning similar things." Consolidate. In refern, you can reassign tags in bulk: select all images tagged with the tag you want to retire, remove it, and apply the canonical tag. Then delete the retired tag.

"I tag diligently for a week, then stop." The workflow is too slow. Set up macros for your three most common image types. Enable the browser extension and use tag-on-save. The goal is tagging each image in under 10 seconds. If it consistently takes longer than that, the taxonomy is too granular.

"I imported from Eagle and my tags did not come over cleanly." refern's Eagle import reads Eagle's folder structure, tags, ratings, sources, and notes. If your Eagle tags used a flat structure, they will arrive as flat tags in refern. After import, open the tag panel and build a parent-child hierarchy by dragging tags onto each other. You can also use the bulk tag editor to promote common tags to the right level.

Next steps

A good tag taxonomy does the heavy lifting, but tags work best as part of a broader organization system.

• What is a reference manager? explains how a persistent library differs from per-project boards and why the two approaches complement each other.
• refern vs PureRef: if you use PureRef for your canvas work, this comparison shows how refern's library and canvas can replace the two-app workflow.
• Best Eagle alternatives for artists: if you are migrating from Eagle and want to understand how refern's tagging compares.

Conclusion

The method in this guide comes down to three decisions made before you tag a single image: design a two-level taxonomy, create macros for your most common image types, and tag at capture time rather than retroactively.

The tool matters because the tag input determines whether you actually use the system. Hierarchical tags with autocomplete and macro support turn tagging from a chore into a 10-second habit. Over time, that habit compounds into a library you can navigate in seconds rather than minutes.