Search Images by Color in Your Desktop Library (2026)

By refern | Last updated: June 2026

The short answer: open the search bar in refern, type color:#c8a87a (or any hex code), and press Enter. refern scores every image in your local library against that target color in milliseconds, fully offline, no internet, no per-search cost. You get a ranked list of images that actually contain that hue, not images you manually tagged with a color label years ago.

This guide explains how that works, when color labels are the better tool, and how color search compares across Eagle, Pixcall, and Cosmos.

Before you start

Color search in refern works on images that have already been indexed and thumbnailed. If you are setting up a new workspace, run the workspace indexer first so refern has time to generate the visual feature data for each file.

You need:

• refern installed and a workspace open (Windows, macOS, or Linux)
• At least one indexed folder with images
• A hex color code for the color you want to find (for example #3a5a8c for a muted navy, or #f2e0c8 for a warm skin tone)

If you do not know the exact hex, use any color picker browser extension, macOS Digital Color Meter, or Windows PowerToys Color Picker to sample a pixel from a reference image.

Step 1. Open the search bar

Click the search bar at the top of your library view, or press Ctrl+K (Windows/Linux) or Cmd+K (macOS) to open the search overlay.

The search bar accepts free text and inline operators at the same time. You can combine color search with other filters in the same query.

Step 2. Enter your hex color using the color: operator

Type color: followed immediately by your hex code:

color:#c8a87a

No space between color: and the hex. The # prefix is optional but recommended to avoid ambiguity.

Press Enter. refern ranks every image in the current scope by how closely its visual profile matches the target color.

Step 3. Read the results

Results are sorted by color match score, highest first. The images at the top contain the most prominent presence of your target color. Images further down still contain the color, but more subtly or in a smaller area.

There is a minimum match threshold. Images that do not contain a meaningful amount of the target color do not appear in the results, so the list stays relevant rather than returning your entire library.

Step 4. Combine color search with other filters

Color search is most useful when combined with other operators. Examples:

• color:#ff6b35 type:image finds only image files (excludes canvases, videos, PDFs) matching the orange
• color:#2d4a6b tag:environment finds environment references that contain that blue
• color:#e8c87a rating:>=4 finds warm-gold images you have rated 4 or 5 stars
• color:#ffffff is:duplicate finds near-white images that may be duplicates

All 14 inline operators in refern can combine freely. The color filter runs server-side in the local SQLite query, so stacked filters do not slow things down.

Step 5. Use the floating color picker for inline search (alternative path)

If you are browsing a folder and want to search by color without opening the full overlay, open the floating toolbar by right-clicking on any empty area in the grid. Switch the toolbar to "search" mode and use the color picker input to enter a hex or click the swatch to open a visual picker.

This keeps you in the current folder context and lets you search within that folder's scope rather than your entire workspace.

Step 6. Narrow with color labels if you want manual control

Color search analyzes pixel data automatically. Color labels are a separate, manual system.

Color labels in refern are one of nine colored dots you assign to an image by hand. They are useful for categorical marking: everything tagged with a red label is "needs review," everything with a green label is "approved," and so on. That is a workflow signal, not a visual analysis.

Use color search when you want to find images that visually contain a particular hue. Use color labels when you want to flag images with a workflow status or a broad category that you define yourself. The two systems complement each other: you can search by color (visual match) and filter by label (workflow state) in the same query.

Common problems and fixes

"My search returned almost nothing." The color match threshold filters out weak matches. Try a color that appears as a dominant or secondary tone in your reference images rather than a tiny accent. If you are looking for skin tones, try sampling directly from the image using a color picker rather than guessing a hex.

"The color I entered matches images I did not expect." Color scoring considers dominant color (weighted at 0.55), HSV histogram (0.35), color layout (0.10), and edge information. A heavily saturated image with a wide color range may score above a more monochromatic image with the target as a small accent. Combine color: with tag: or type: to narrow the results.

"My images were not re-indexed after I edited them." refern detects file changes on disk through the file watcher and the reconcile/resync workflow. After editing an image in an external editor, use the "Resync" option (the toast that appears when file changes are detected) or run "Sync with disk" from the workspace dashboard to regenerate the thumbnail and visual features for changed files.

"I want to search by color across multiple workspaces." Color search runs within the currently open workspace. To search across workspaces, switch workspaces from the workspace selector. Cross-workspace search is not yet supported.

How color search compares across tools

Color search is available in Eagle, Pixcall, Cosmos, and refern, but the implementations differ in meaningful ways.

Eagle has a mature, well-regarded color search. The adjustable accuracy slider is useful for broad-vs-narrow matching. It runs locally on your Eagle library. Eagle's library requires importing files (which copies them into a proprietary .library folder), so you are searching Eagle's copy of your files rather than your originals [eagle.cool, AlternativeTo, as of 2026].

Pixcall extracts a 9-color palette per image. A V2EX user testing Pixcall in 2022 noted the color filter was "too precise" (requiring exact hex values) and that multi-color filtering was additive with no toggle, which created confusion [V2EX, 2022]. Pixcall's color filtering has developed since that report, but the 9-palette model differs from the full-histogram approach.

Cosmos provides hex color search that is specifically highlighted as a standout feature for designers [Medium review, cosmos.so, as of 2026]. It works well for the use case Cosmos is built for: searching web-saved inspiration in your cloud collection. It does not help you search local files on your disk, and it requires an internet connection.

refern runs color search against images already indexed in your local library, fully offline, with no API, no cost per search, and no upload of your files anywhere. The color: operator is composable with all 13 other inline operators, so you can layer color with tags, ratings, file types, dates, and relationship filters in a single query.

Why "search images by color" and "color label images" are different problems

A common question is whether to use color search or color labels. They solve different problems.

Color labeling is a manual workflow tool. You assign a label to communicate something about your process: this image is approved, this one needs editing, this one belongs to a warm-palette set. It is fast to filter and useful for team conventions (if you are doing code reviews or a mood board review with collaborators). The limitation is scale: labeling thousands of images by hand is impractical.

Color search is automatic. refern computes a visual feature vector for every image at index time. When you type color:#3b6ea5, the query scores every image in your library and returns those that actually contain that blue, ranked by how much and how prominently. You do not need to have labeled anything in advance. This makes it practical at any scale.

Use color labels to communicate workflow state and manual categorization. Use color search to find images by what they actually look like. Use both together to narrow a visual search to a pre-labeled subset (for example: images you've marked as approved that also contain a warm sunset palette).

Next steps

Once you have found images by color, you have several natural next steps:

• Bring them onto a canvas. Drag color-matched images from the search results onto an infinite canvas to arrange them into a palette study or mood board. See the canvas documentation at refern.app.
• Tag them for future filtering. Select the results and apply a tag so you can quickly re-filter this set without running the color search again.
• Find visually similar images. Right-click any image and choose "Find similar" to run a broader visual similarity search (beyond color alone) that compares structure, edge distribution, and dominant tones.
• Compare with Eagle's color search. If you are coming from Eagle and evaluating refern, see refern vs Eagle for a full feature comparison.
• Explore other search operators. refern supports 14 inline operators including tag:, rating:>=3, type:image, is:duplicate, derived:, and linked:. All can combine freely in the search bar.