重複ファイルの整理 alternatives: what you can drop

A duplicate cleaner is an easy thing to buy and a hard thing to justify twice. The first run feels productive, several gigabytes come back, and then the license sits unused for eleven months until the drive fills again. That gap between purchase and use is what sends people looking for an alternative, and the honest answer depends less on which app is best than on which parts of the job macOS already performs for free.

The job splits cleanly. Matching is arithmetic. Deciding is not. Almost every alternative competes on the arithmetic.

Photos already ships with this, and it understands the library

If the duplicates causing trouble are photographs, the decision is short. The Photos app has duplicate detection built in, listed as Duplicates under Utilities in the sidebar. If the entry is not visible, hovering over Utilities reveals a button that adds it. Individual rows can be merged, or Command-A selects everything at once for a single bulk merge. Anyone on an iCloud Shared Photo Library should set the Library menu in the toolbar to Both Libraries first, or half the duplicates stay hidden.

Merging leaves one original in place. The copies it absorbed go somewhere recoverable.

Deleted duplicates appear in Recently Deleted, where you can recover them within 30 days or permanently delete them. Source: support.apple.com

The reason to prefer this over any external scanner is structural, not financial. A photo library is a managed container with albums, edit history, and keyword metadata pointing at specific assets. A general purpose scanner walking the same folder sees files and nothing else, and removing one can break references the library still holds. Apple's own tool merges at the library level, so the album memberships and edits follow the survivor.

The limit is scope. It only looks inside the library. Exported JPEGs scattered across Downloads and Desktop are a separate problem, handled below.

The matching is arithmetic, and the arithmetic ships with macOS

The part that paid apps advertise most loudly is the part that is least scarce. Identifying byte identical files means computing a digest for each candidate and grouping by the result. macOS includes the tools for that in the base system, with no installation and no license.

The practical shape is: list the files, compute a digest per file, sort, and keep the groups that appear more than once. Restricting the first pass to files of identical size cuts the work dramatically, because two files of different sizes can never be byte identical, and digests are only worth computing inside size groups that already collide.

What this replaces is not a convenience. It is a correctness check. Matching on filename and size alone produces false positives that look extremely convincing: two exports from the same camera at the same settings can land on identical byte counts with different content, and two entirely different documents can share a name across folders. Any tool that offers name based matching as a default deserves a spot check, and running a digest comparison on a handful of its proposed pairs is the cheapest possible audit.

The output has a second use. A digest list is a text file, which means it can be diffed against a later run, committed somewhere, or reviewed by someone else before anything gets deleted. No graphical scanner produces an artifact that survives the session in that way unless it offers an export.

Free tools cover the same ground

For anyone who wants a list without writing one, the open source options match commercial scanners on the matching stage. dupeGuru is the common recommendation: it is free, currently at version 4.3.1, installs through Homebrew as a cask, and runs natively on Apple Silicon. It offers selectable match modes rather than a single opaque algorithm, which is the property that matters when the results need to be trusted.

Where free tools tend to be weaker is the interface for reviewing thousands of matched pairs, and in the polish of the selection helpers. That weakness is worth examining rather than assuming, because the selection helpers are on the list of things to drop.

Installation route matters slightly here. A cask install through a package manager records what was installed and removes cleanly later, which is a small advantage over a downloaded disk image for a tool that only runs occasionally. The same applies to the command line deduplication utilities available through the same channel, which produce the same digest based grouping described above without any interface at all.

A file manager may already cover the scoped version

There is a middle option that gets skipped. Scanning an entire drive is a different task from checking one project folder, and the second is far more common. Cleaning up a client folder before archiving it, or verifying that a card import did not land twice, needs a scan with a boundary, not a scan of everything.

Several file managers now include exactly that. Path Finder 26, for instance, added the ability to find duplicated files within selected folders, alongside an operation for comparing two folders side by side. No additional purchase, and the scope is set by whatever is already highlighted.

Folder comparison deserves separate mention, because a large share of what gets called a duplicate problem is actually two versions of the same directory rather than scattered individual files. An old backup copy of a project sitting next to the live one produces hundreds of duplicate pairs in a flat scanner, which is hundreds of individual decisions. Resolved as folders, it is one decision. The way tools differ on scope and on folder level operations is laid out in Compared with other file managers.

Three features that can go

Paid scanners differentiate on capabilities that sound necessary and mostly are not.

Whole disk scanning. Pointing a scanner at the entire volume drags in application bundles, package manager caches, version control working trees, and system support directories, all of which contain legitimately identical files that exist for a reason. The result is a list dominated by matches nobody should act on. A scan scoped to Downloads, Desktop, Documents and one external drive finds the duplicates a person actually created.

Similar file matching. Detecting near identical images, such as burst shots or the same photo at two resolutions, is genuinely difficult and genuinely a paid feature. It is also a category where automated judgement is weakest, because the difference between a redundant copy and a deliberately kept variant is not visible in the pixels. For documents and project files, this mode contributes almost nothing.

Scheduled scans. Running a scan every week without closing whatever keeps producing copies just converts a one time cleanup into a recurring chore. The sources are usually identifiable: a download that gets re run rather than located, an export step that writes to two places, a sync conflict pattern, a backup folder nested inside a working folder. Fixing one source removes more future work than any schedule.

Two features that should not go

An exclusion list. This is the single most important capability to verify before switching tools, and the easiest to overlook because nobody markets it. The scanner needs to be told, explicitly, where not to go: the Photos library, application bundles, anything under a version control directory, dependency folders, Time Machine volumes, and any managed container belonging to another application. A tool without exclusions is not cheaper, it is riskier, and the risk lands on exactly the directories where deletion does invisible damage.

An export of the deletion list before deletion. Something must be able to write out what is about to be removed, as plain text, so it can be read before the operation runs and referenced after. Trash provides recovery, but recovery requires knowing what to recover. A list of 400 paths in a text file answers that question. A dialog reporting that 400 items were cleaned does not.

	Photos app	Digest comparison	Free scanner	File manager built in	Paid scanner
Scope	Photo library only	Anything specified	Chosen folders	Selected folders	Whole volumes
Match method	Library aware	Byte identical	Selectable modes	Exact match	Includes similar files
Selection help	Merge action	None	Manual review	Manual review	Automatic selection
Cost	Included in macOS	Included in macOS	Free	Included in the app	Subscription or one time

Comparing the cost shapes, not just the prices

Cleanup is episodic, which makes the billing model matter more than the headline number. Gemini 2 from MacPaw, one of the established options in this category, lists pricing starting at 1.95 USD per month and also sells monthly, annual and one time plans covering one, two or five Macs. For a once a year cleanup, a single month followed by cancellation costs least. For continuous use, annual wins. For the same machine over several years, the one time license does.

Bundles change the arithmetic again. Setapp includes a large catalogue for 14.99 USD per month plus tax on its Mac plan, with a 7 day free trial, and access ends when membership ends. That is not a comparison against one cleaner's price. It is a comparison against the other applications in the bundle that would genuinely get used.

The free route costs time instead of money: reading the matching modes, setting exclusions, and building a review step. An hour spent once is a fair trade against a recurring subscription for a task performed twice a year. It is a poor trade for someone cleaning client deliverables every week. How these cost shapes line up across tools is set out on Pricing.

Where the time actually goes

Every alternative above competes on producing the list. None of them shortens what happens next. Given 500 matched pairs, some are a file and its backup, some are a deliverable and the working copy it came from, and some are last quarter's invoice template and this quarter's. Size and timestamp do not resolve any of those, because the newer file is frequently the derivative and the older one is frequently the original.

Resolving a list like that means moving between a listing, a preview of the contents, and the place where the removal happens. Three contexts, several hundred times. The scan took ninety seconds. The review takes the afternoon.

That is the part worth measuring before buying anything. If the scan is the bottleneck, better matching helps. If the review is the bottleneck, and it usually is, a faster scanner produces the same list sooner and changes nothing about the afternoon.

What to change first

Run the Photos duplicates view today, since it costs nothing and removes an entire category from the problem. Then scope the next scan to the folders a person actually created files in, and set exclusions before running anything. If the remaining time goes into working out what each file was for rather than finding it, that is a different problem from duplication, and Atriens is built around it.