Find My Old Reference Images Organized by Tag (2026)

You know you saved it. You remember the mood, maybe the color palette, maybe the site you found it on. But you cannot find it now. This is the most common pain in reference management, and it is almost entirely caused by the tools, not by you.

Most reference tools were designed for one of two things: floating images on your screen while you paint, or sharing inspiration boards with clients. Neither use case requires a real search engine. The result is that PureRef, BeeRef, Milanote, Pinterest, and Are.na all make retrieval from a large personal collection extremely difficult or outright impossible.

refern is a local desktop reference manager built specifically to solve this. Full-text FTS5 search, 21 user-visible operator keywords, color search, image-to-image visual similarity, and duplicate detection run on the device.

Why You Cannot Find That Image (and Which Tool Is at Fault)

PureRef: no search exists at all

PureRef is the most widely used reference overlay tool in game art and concept art studios. It is excellent at putting images on your screen while you work in another app. But the official PureRef handbook states plainly that there is no way to search for an image by name, keyword, tag, color, or any other attribute within PureRef. A 2022 thread on the PureRef forum explicitly requested folders, tags, and tabs for organizing references. That feature has not shipped as of PureRef v2.1.3 in June 2026.

If your board has 400 images and you need a specific hand anatomy study from six months ago, your only option is to scroll and scan the canvas visually. This is not a workflow; it is a lottery.

PureRef also creates a structural retrieval problem: each .pur file is a self-contained board. There is no cross-project library. A reference in a "character study" board from March is invisible when you are working on an "environment art" board in November. You have to remember which file it is in, open it, and scroll.

PureRef is genuinely excellent at what it does. Using it as a library is the problem.

BeeRef: same limitation, free price

BeeRef is a free, open-source alternative to PureRef built in Python and licensed under GPL-3.0. It covers the basic canvas overlay use case well and has no cost. But it has the same retrieval ceiling: no tags, no search, and no metadata of any kind. The BeeRef homepage confirms this directly.

AlternativeTo users consistently list the lack of tags and search as BeeRef's primary limitation. Each .bee file is one self-contained scene. A user with 200 saved scenes has no in-app way to browse, preview, or search them. The file format does not produce OS-level thumbnails, so even the file manager gives you no visual preview without opening each one.

BeeRef is a solid free tool for "float a few images over my painting app during a session." It was not designed as a library.

Milanote: keyword search that misses your images

Milanote is a polished, cloud-based visual workspace with genuine strengths in collaboration, client presentation, and templates. Its search, however, is limited to basic keyword lookup over card titles. If you added an image to a board without a title, that image is effectively invisible to search. There is no tag system, no color search, no visual similarity, and no operator-based filtering.

A GetGuru analysis of Milanote's search notes that "users often struggle to find specific notes or items" and that the system has no adaptability to synonyms or phrasing variation. A 2026 App Store review complained that a recent update "limits search to recently viewed or edited content, making finding specific boards a slow, frustrating experience."

Milanote also hits a performance wall around 300 to 500 images per board: lag and crashes become common at that scale. It is a strong tool for presenting work to clients. It was not designed to manage a long-term growing image library.

Pinterest: your Pins are not really searchable

Pinterest is where many artists begin collecting references. The discovery feed is genuinely powerful for finding new content. Your personal saved library, however, is nearly impossible to search systematically.

Pinterest search covers only Pin titles and Board titles. There are no inline operators, no rating filters, no color search, and no visual similarity search over your personal collection. On April 29, 2026, Pinterest's search experienced a partial outage affecting users across iOS, Android, and web simultaneously.

There is also a structural impermanence problem. Pins are links to external URLs. When a source page is deleted, the reference link breaks. Images you saved two years ago may now point to nothing.

Note: refern is a local desktop tool you own, not a hosted social feed. If you want Pinterest's discovery experience, refern is a different kind of tool. If you want to own and search the references you have already collected, refern was built for exactly that.

Are.na: community curation without content search

Are.na is a thoughtful, ad-free bookmarking platform built around the concept of "blocks" that can live in multiple "channels" simultaneously. Its no-algorithm ethos and collaborative channels attract designers, academics, and artists who want intentional curation over an algorithmic feed.

But Are.na's search covers channel names and usernames, not the content of blocks. A user with hundreds of channels has no way to find a specific reference image by color, content, or metadata. Are.na's strength is community and collaboration. It was not designed as a searchable local image library.

What a Searchable Reference Library Actually Requires

To find an image you saved months ago with confidence, you need four search layers working together.

Most tools have zero of these layers. A truly searchable reference library has all four.

How refern Solves the Retrieval Problem

refern is a desktop reference manager for artists that combines Eagle-style organization with a PureRef-style infinite canvas and an Obsidian-style relationship graph. It costs $35 one time and runs on Windows, macOS, and Linux. Normal workspace indexing reads existing source files in place. Its SQLite index and generated thumbnails live alongside, or in a configured data location associated with, the workspace.

Full-text search with 21 user-visible operator keywords

Every image in refern has a searchable record covering filename, path, description, notes, source URL, creator, and all tags. SQLite's FTS5 engine uses BM25 ranking to surface the most relevant results first.

You combine this with inline operators:

• tag:anatomy to find everything you tagged with that term
• rating:>=4 to surface only your high-quality saves
• type:image to filter to images only, excluding canvases and other file types
• source:artstation.com to find everything from a specific origin
• is:duplicate to surface images you may have saved twice
• linked:true to find images connected to other images or canvases via typed entity links
• sort:added or sort:rating to control result ordering

These operators compose freely. A query like tag:character rating:>=3 source:artstation.com returns high-rated character references from ArtStation in one pass.

Hierarchical tags

Tags in refern nest. You can build a structure like Character > Anatomy > Hands and then search tag:Hands to find every image at that level, or tag:Anatomy to retrieve everything in the whole subtree.

Tag groups cluster related tags for easier browsing. Linked tags create automatic associations (tagging one image Hands can auto-suggest Anatomy). Tag macros let you apply a bundle of tags from a single typed shortcut, which is useful for consistent tagging across a session.

When you import an image, refern reads any embedded EXIF, IPTC, or XMP metadata automatically. If a photographer tagged an image in Lightroom or digiKam, those tags carry over without manual work.

Color search

If you remember "it was that warm copper-toned portrait study" but nothing else, color search can help narrow it down. Paste a hex code or click a color swatch in refern's search bar. The search processing runs on the device without uploading the library to a refern cloud service.

Visual similarity search

If you have one image and want to find everything in your library that looks similar, visual similarity builds a local 512-byte descriptor from each image covering HSV histogram, dominant colors, color layout, and edge histogram, then runs a weighted comparison. Right-click any image and choose "Find similar" to get results ranked by visual closeness across your whole library.

Visual similarity processing runs on the device and does not require uploading the library to a refern cloud service. This is a core-workflow boundary, not a claim that the whole application makes zero network requests: activation, updates, optional model downloads, feedback, and user-selected web downloads use the network.

Cross-project library

Unlike PureRef's per-.pur boards or BeeRef's per-.bee scenes, refern's workspace covers your entire folder structure. Every image imported across every project is indexed in one place. A character study you added in March is findable from a November search with no manual file-switching.

Smart folders extend this further. A smart folder is a saved search query that auto-populates as your library grows. Create one for tag:anatomy rating:>=4 and it always surfaces your best anatomy references, across all projects, in real time.

Quick Comparison: Search Across Tools

How to Set Up Your Library So You Can Always Find Things

Use directory metadata presets

In refern, you can attach a metadata preset to any folder. When a file lands in that folder by import, drag-drop, or browser extension save, the preset applies its tags automatically. A folder named "Anatomy References" can tag every incoming file with anatomy and assign a color label without any manual action.

Set it up once per folder. It runs on every future import.

Save source URL at import time

The browser extension (Chrome, Firefox, and Safari) captures the source URL of every image you save. When you search source:artstation.com six months later, every ArtStation reference surfaces immediately regardless of how you tagged it. This also matters for attribution: refern retains the source domain, the page title, and the EXIF creator field, so you can always trace where a reference came from.

Use color labels for visual triage

refern has nine color labels you can assign to any image. A simple scheme: red for "low quality or do not use," green for "strong reference," yellow for "context needed." Then filter with colorLabel:green when you need only your best saves for an active project.

Run duplicate detection periodically

As your library grows across months and projects, you will save the same image twice from different sources. The is:duplicate operator uses perceptual hash comparison to find images with the same visual content regardless of filename or source URL. Running it once a month keeps your library clean.

Common Problems and Fixes

Problem: I organized images in PureRef notes but now they are gone. PureRef has no tags. Floating text notes on the canvas are spatial objects, not searchable metadata attached to the image files. They are not recoverable as a tag structure. If you want to preserve the information, open the PureRef board, read each note manually, and re-enter the terms as tags in refern after import.

Problem: Images I organized in Milanote boards are now impossible to find. Milanote boards are not a searchable library. Images without explicit card titles are invisible to Milanote's search. Your options are to manually add titles to each image card inside Milanote, or export images as a ZIP and import them into refern, where FTS5 search and color search will immediately make them findable.

Problem: My Pinterest reference images are broken because the source pages disappeared. Pinterest Pins are links. When the source page is gone, the image may be inaccessible. If the thumbnail is still cached, right-click and save it, then import into refern. Going forward, refern's browser extension saves the actual image file to disk at import time, so the reference persists even if the original URL disappears later.

Problem: Are.na search returns channel names, not the image I am looking for. Are.na has no block-level content search. Your options inside Are.na are to browse channels manually or use browser Ctrl+F on a channel page. If this is a recurring problem, moving your working image library to a tool with real search is the long-term fix. The guide on how to organize reference images covers the options.

Honest Gaps

refern is a local-first desktop app. It has no cloud sync, mobile app, or collaborative sharing today. Cloud sync and shareable canvases are planned without published dates. If you need to share a reference board with a client who will not install a desktop app, Milanote is the better tool for that specific case today. If you primarily want to discover new references you have never seen before, Pinterest or Are.na serve that use case better than refern.

refern's 30-day trial is fully functional with no account required, so you can test whether the search actually works for your library before paying anything.

Next Steps

If the "I know I saved it" problem is familiar, the underlying cause is almost always the tool. A reference collection that cannot be searched is a liability, not an asset.

• Compare search capabilities in depth on the refern vs PureRef comparison page
• Learn how tagging works alongside folders in folders vs tags for reference images
• If you are moving away from PureRef, read how to migrate from PureRef to refern
• Start from the basics with what is a reference manager