Duplicate File Finder

Connect folder asks the browser for read access. Browsers that support the folder picker use it directly; otherwise the app falls back to folder selection through a file input.

1. Choose Connect folder.
2. Select the folder you want to review.
3. Wait while File Finder builds a local in-memory index.
4. Open Duplicates first if your goal is duplicate cleanup planning.

Try guided demo loads fictional files so you can practice without connecting your own folder.

Exact copies

Exact-copy groups require matching file size and matching SHA-256 content hash. Same-size files of up to 128 MiB each are eligible for verification in this build. Empty files are excluded from duplicate detection.

Possible copies

Possible copies are candidates that do not have enough content-verification evidence to be called exact. Treat them as a review queue, not proof.

Name conflicts

Files can share a filename and still contain different bytes. The guided demo includes this case deliberately.

ZIP files are compared as whole files. File Finder does not claim that two different ZIP containers are duplicates merely because their unpacked contents might resemble each other.

Choose Review copies on an exact group. Each copy card shows the context that can differ even when file bytes are identical: filename, folder, size, extension, and modified date.

• Open in Viewer previews supported files read-only.
• Copy path copies the displayed path so you can locate the file yourself.
• Select two copy cards to compare their context side by side.
• Expand verification or technical details only when you need deeper evidence.

The review plan is a session-only decision aid. Available labels are Keep in plan, Potential redundant copy, Review later, and Not labeled.

An exact group counts as reviewed only when every copy has a settled label and at least one copy is marked Keep. Keeping every copy is valid.

Download the readable Markdown plan or JSON plan before reloading if you want to retain your decisions. Exports contain review information and paths; they do not contain executable cleanup commands and do not modify source files.

Files is both the browser and search workspace. With an empty query it shows the connected index. Type a query and the same table becomes search results.

Filters, sorting, and extra columns live behind the options control so the default screen stays calm. More search tools contains examples, saved searches, facets, and advanced syntax such as ext:zip, is:duplicate, project:"Client A", or -ext:zip.

Words inside supported files become searchable only after you explicitly run Analyze file contents in Inside Files.

Open files from Files, Duplicates, or other workspaces. Viewer can preview supported images, PDF/audio/video, readable text, Markdown, JSON, delimited tables, and package signatures/listings depending on file type.

HTML and SVG are treated cautiously: active document content is not executed by the Viewer. Unsupported files still retain their indexed metadata and relationships.

Review collects strong local signals from analyses the app has already run: exact copies, possible copies, version overlap, connection problems, and changes. Areas can overlap, so counts are not a count of unique “bad files.”

No review signal means only that nothing is currently visible from the analyses already performed. It does not prove a folder is clean, complete, or correct.

• Storage shows what uses space and where, including duplicate overhead evidence.
• Activity organizes file modified times and dates found inside files.
• File Map visualizes files, folders, projects, duplicates, versions, and connections.
• Inside Files performs optional local content analysis on supported readable files.
• Versions groups likely related versions without silently choosing an authoritative file.
• Connections shows explicit local links and recognized relationship metadata.
• Changes compares the current folder against a local baseline snapshot.

Click a node to open its Node Card. Double-click a node to center and focus it. Focus can limit the graph to one- or two-step neighborhoods so large maps stay understandable.

Use semantic zoom: overview levels show less labeling; closer zoom reveals richer file labels and type cues. Empty-space click or Escape clears selection.

The graph is an exploration aid. Relationship edges represent the evidence types File Finder knows; they do not imply that one file should be deleted.

Last modified is the timestamp supplied by the browser for the selected file. It is not necessarily the date a file was originally created or copied.

A creation or updated date is shown only when File Finder actually finds a recognized date field inside file metadata. When that evidence is absent, the app says the creation date is unavailable rather than inventing one.

Content analysis is explicit. It indexes supported readable content locally so Files search can find words, headings, and locally derived terms inside documents.

This build limits readable-content analysis to up to 2 MB per file and 64 MB per indexing run. The content index stays in memory for the connected session and can be cleared without changing files.

Storage provides folder, project, and type views of the connected index. Duplicate overhead is derived from extra members of exact-copy groups and is evidence for review, not a deletion instruction.

Click items to drill or inspect. Large visualizations may summarize display while the connected file index remains available elsewhere in the app.

File reading and analysis are designed to run in the browser. The tool does not require network requests for its file-analysis workflow and does not upload your selected files.

Review labels and content indexes are session data. Saved-search definitions can use browser local storage, but the saved-search feature stores the search definition rather than your file contents.

Exports are newly created local files. Review them before sharing because they can contain filenames, paths, or other information from your workspace.

• Ctrl/Command + K · open Find & Commands.
• / · open Files and focus file search when you are not typing in a form field.
• Alt + Left / Right · move through the app’s view history.
• File Map: arrow keys pan, +/- zoom, F fits, 0 resets, and Escape clears selection.

Folder does not open

Check that your browser allows folder/file selection and that you granted access. If the modern folder picker is unavailable, the app attempts its folder-selection fallback.

A file is only a possible copy

Verification may be unavailable or intentionally bounded. Large files above the exact-hash limit remain clearly separated from verified exact copies.

No duplicates appear

This can mean no exact groups were verified, but it can also reflect the current scan/verification limits. Check Possible copies and make sure scanning completed.

Preview unavailable

The file remains indexed. Use its metadata, path, duplicate context, or external application if the browser cannot preview that format.

Large folder feels busy

Open Folder details on Home to review responsiveness controls. These controls change scheduling/display limits, not the completeness of Files search.

• SHA-256 / hash: a content fingerprint used here to verify exact byte matches.
• Exact copy: files with matching size and verified matching content hash.
• Possible copy: a candidate that lacks enough verification evidence to be called exact.
• Metadata: structured information associated with a file or found inside supported file content.
• Snapshot/baseline: a local reference used by Changes to compare a later folder state.
• Node: an item shown on File Map, such as a file, folder, project, duplicate group, or relationship entity.
• Relationship: an explicit or recognized connection between local items; not an instruction to modify them.