Introducing AI Search — Visual Search & Natural Language Search, Fully Offline

Inside EagleEagle Team2026-03-17

Today, we're officially launching AI Search, an Eagle plugin built for a smarter way to find your assets.

With this release, AI Search brings two new ways to find assets in Eagle:

Reverse Image Search: Find visually similar assets using a reference image
Natural Language Search: Search through your assets' text information using plain descriptions
Fully Local: No internet connection required, no uploads — works completely offline

For anyone who has built up a large asset library, AI Search isn't just a new feature — it's a fundamentally more intuitive way to find what you're looking for. Whether you start with a reference image or type out a description, it gets you closer to how you actually think about and recall your assets.

Why We Built AI Search

Once your library grows into the tens of thousands of images — or even hundreds of thousands — folders, filenames, and tags can only take you so far.

You've probably run into situations like these:

You remember saving an image, but the filename is long gone from memory
You have a reference image and want to find assets with a similar look or feel
You're searching for a certain type of subject, but can't think of the right keywords
Your library keeps growing, and traditional search just isn't keeping up

AI Search was built specifically to address these pain points.

It's not meant to replace folders, tags, or keyword search — it's meant to complement them with two more intuitive modes: reverse image search and natural language search.

Reverse Image Search: Find Similar Assets with a Reference Image

With reverse image search, you can drag in any image and let AI Search find visually similar results from your library.

Search results are based on the image's overall visual characteristics, including:

Color palette and tones
Composition and layout
Visual style
Subject matter and overall feel

This type of search is especially useful for things that are hard to put into words. A subtle color mood, a particular layout rhythm, or the overall atmosphere of an image — these are difficult to capture in a few keywords, but a reference image can get you there much faster.

How to Use It

There are two ways to start a reverse image search:

Using an image already in Eagle

Drag any image from the file list directly onto the reverse image search button next to the search bar, then release to begin searching.
Using an external image

Copy an image to your clipboard, then click the reverse image search button and paste — it'll use that external image as your reference.

Where to Find It

You can access reverse image search from the following locations:

To the right of the search bar
Pinned to your filter bar

Click the "+" in the filter bar to add reverse image search to your quick-access area.

Tips for Best Results

If you have a general sense of what you're looking for but aren't sure which specific image, try starting with a keyword or natural language search to narrow things down. Once you find something close, use that as your reference image to expand the search from there.

This "find one, then branch out" approach tends to be significantly more efficient when working with large libraries.

Natural Language Search: A More Flexible Way to Search Your Text Metadata

In addition to reverse image search, AI Search also supports natural language search.

This feature currently works by performing semantic matching against the text information already associated with your assets, such as:

Filenames
Descriptions
Notes
Other searchable text fields

Think of it as semantic text search: a more flexible way to match against filenames, descriptions, and notes — rather than analyzing image content directly through visual understanding.

Unlike traditional keyword search, which matches text literally, natural language search attempts to understand semantic proximity. So even if the words you type don't exactly match how an asset was originally named or described, results with a close enough meaning can still surface.

For example, if your assets have reasonably clear names or descriptions, you can search with phrases like:

warm-toned minimalist poster
flat style animal illustration
クリスマス
dark moody portrait

and find matching content.

Natural language search works especially well when:

You remember a theme or concept, not an exact keyword
Your assets already have filenames, descriptions, or notes
You want a more intuitive way to search through a well-organized library
You need to search across multiple languages within your existing text metadata

How to Enable It

Click the "+" in the filter bar
Select "Natural Language Search"
Type your description or keyword in the input field

You can also pin it to your quick-access area for easy access later.

Before You Start: Where AI Search Works Best

We want to be upfront about how each search mode works.

The two search modes in AI Search are built on fundamentally different foundations:

Reverse Image Search: Based on visual similarity between images
Natural Language Search: Based on the text information already attached to your assets

This means natural language search tends to shine in situations like:

Filenames are meaningful and descriptive
Descriptions or notes are filled in clearly
Your library already follows consistent organizational habits
You need to handle synonyms, cross-language queries, or varied ways of describing the same thing

On the other hand, if an image has no meaningful text associated with it — a random filename, no description, no notes — there's simply less for natural language search to work with, and results will reflect that.

This isn't a limitation of the search engine itself. Even the most powerful search needs something meaningful to match against.

Fully Local — No Internet, No Uploads

Every step of AI Search's analysis and retrieval happens on your own machine.

That means:

Your images are never uploaded to the cloud
No external servers involved in the analysis
Works even when you're offline
Your assets never leave your device

Feature	Other AI Search Tools	Eagle AI Search
How it works	Upload to cloud → Analyze → Return results	Analyze locally → Search locally
Privacy	Assets pass through third-party servers	Assets stay on your device
Cost	Usage-based billing / subscription	Free to use after installing the plugin
Offline use	Requires internet connection	Fully offline capable

For anyone managing client design files, a personal portfolio, or an internal reference library, this local-first approach fits more naturally into existing workflows and is much easier to sustain long-term.

GPU Acceleration for Large-Scale Processing

AI Search supports GPU acceleration. On compatible hardware, overall processing speed can be up to 30–50x faster than CPU-only mode.

Environment	Support
Windows + NVIDIA GPU	Supports both GPU and CPU modes — GPU recommended
Windows + AMD / Intel GPU	CPU mode only (currently)
Mac (Apple Silicon)	GPU acceleration supported (enabled by default)
Mac (Intel)	CPU mode only

If you have a large library, or frequently run indexing, batch analysis, or repeated searches, the difference with GPU acceleration will be very noticeable.

If you're limited to CPU mode, AI Search will still run — but initial indexing of large libraries or batch image analysis will take considerably longer.

The setup wizard automatically detects your hardware and recommends the right version. No manual configuration needed.

Who Is AI Search For?

AI Search is particularly well-suited for users who:

Have built up a large library and need more efficient ways to find things
Remember assets visually rather than by filename
Already have consistent naming conventions, descriptions, or notes in their library
Want to pair it with AI Action in the future, creating a complete organize-and-search workflow

The larger your library, the more significant the difference AI Search tends to make.

What's Next: AI Action

AI Search addresses the question of how to find your assets.

Our next AI plugin feature, AI Action, tackles the complementary question: how to make your assets easier to find in the first place.

AI Action focuses on batch organization tasks, such as:

Auto-naming
Auto-generating image descriptions
Auto-tagging
Auto-categorizing

For us, the value isn't just generating generic captions. It's about letting you define your own organizational logic — how assets should be named, described, and categorized — and then having AI execute that at scale.

That's what makes AI Search and AI Action so powerful together:

Use AI Action to enrich your assets' text metadata

Build clearer names, descriptions, and tags across your library
Then use AI Search to find them

Whether you're searching by description or branching out from a reference image, everything becomes more efficient

AI Action is currently in its final testing phase. We'll share a dedicated introduction soon.

A Note on Eagle 5.0

Sharp-eyed users may have noticed that we haven't bundled any of these new features under the Eagle 5.0 banner. That was a deliberate decision made after reassessing our approach.

We had originally planned to ship these capabilities as part of an Eagle 5.0 Beta. But we came to believe that requiring users to upgrade to a major new version — with all the underlying changes that entails — just to access these features wasn't the right call. Instead, we chose to deliver them on the stable foundation of Eagle 4.0, so more people can benefit right away.

The reasoning is straightforward: we didn't want users to have to absorb the compatibility, stability, and migration risks of a major version upgrade just to try out AI Search and AI Action. Getting new features into users' hands in a stable, accessible way matters more to us than advancing the version number.

That said, Eagle 5.0 is genuinely in development. For us, 5.0 isn't just about adding new features — it's about upgrading Eagle itself: core runtime updates, UI improvements, performance optimizations, and fixes for issues that have been difficult to address. We expect 5.0 to bring meaningful performance gains across the board.

One concrete example: after the underlying runtime is updated, Eagle 5.0 will require macOS 12 or later as the minimum supported version. That's exactly why we didn't want to tie new features to that transition — we didn't want users still on older hardware or older operating systems to be shut out before they're ready.

In the near term, Eagle 4.0 will continue to be maintained, with features we've committed to being rolled out progressively via plugins. Looking further ahead, Eagle 5.0 remains our primary development focus — a more comprehensive upgrade to the application itself.

Looking Further Ahead: Eagle Agent

Beyond search and organization, we're also thinking about more natural ways to interact with AI directly inside Eagle.

We recently launched Eagle MCP/Skill, which lets external AI assistants collaborate with Eagle. Our next area of exploration is a step further: Eagle Agent.

Our goal is to bring conversational AI natively into Eagle — so users can work with AI naturally, without needing to set up external tools, and accomplish more within the asset management workflow itself.

This direction is still in early stages, but we're genuinely excited about where it could go.

Get Started Today

Open Eagle, head to the Plugin Center, and search for AI Search to install it.

If you regularly deal with large asset libraries, reference image collections, or inspiration searches, we'd love for you to give it a try. For us, this isn't just a new feature release — it's a meaningful step forward in how Eagle helps you find what you're looking for. We're genuinely curious to hear whether it makes a difference for you.