How does Duplicate Photos find WhatsApp and resized photos when file sizes and hashes differ?

WhatsApp and messaging apps re-encode photos into lower resolutions with heavy compression, completely changing their file sizes and byte hashes. Duplicate Photos uses a 4-tier engine: it computes difference hashes (dHash) and discrete cosine transform perceptual hashes (pHash) indexed in a metric BK-Tree. When candidates are found, it runs a 128x128 Pearson pixel correlation test. If the visual agreement across scales remains at r >= 0.99, it identifies the image as a true re-encode and automatically preserves the highest-resolution camera original.

Why won't Duplicate Photos accidentally delete my burst photos of people moving?

Burst shots (e.g. 5 photos taken in 1 second) share similar overall color and 64-bit hashes, tricking crude duplicate finders into grouping them together. Duplicate Photos evaluates multi-scale pixel correlation: true re-compressed copies increase in mathematical agreement as resolution rises (from 0.9980 at 32x32 to 0.9999 at 256x256), whereas burst frames diverge in micro-detail (from 0.9928 at 32x32 down to 0.9628 at 256x256). The engine rejects 98.6% of burst shots and routes them into the 'Similar Photos' view where no files are auto-selected.

Is it completely safe? What if I accidentally clean something I wanted?

Safety is hard-coded into the Rust core engine. First, removals are sent to the Windows Recycle Bin using native Win32 Shell API rather than unlinked permanently. Second, every cleanup session is indexed in a local history log and can be reversed with a single click. Third, the engine enforces a star-topology invariant: selecting all items in a group will still hold back the primary Keeper copy, making it impossible to delete all instances of a photo.

Are my photos uploaded to any server or cloud AI?

No. Duplicate Photos is 100% local and offline. All image decoding, hashing, BK-tree clustering, and pixel correlation happen entirely in Rust on your local CPU. The app requires no account, displays no advertisements, and never transmits your photos, filenames, or metadata over the internet.

How does it handle Live Photos, RAW files, and cloud storage folders?

Duplicate Photos is companion-file aware: iPhone Live Photo clips (.MOV) and RAW image sidecars (.XMP / .AAE) are linked to their parent photo and moved together so files are never orphaned. Additionally, it inspects Windows file system attributes (Reparse Points / Sparse Files) to detect OneDrive, Dropbox, and iCloud offline placeholders, skipping them automatically so your computer doesn't trigger unexpected cloud downloads.

How much does Duplicate Photos cost and what are the system requirements?

Duplicate Photos is 100% free for Windows 10 and Windows 11 (64-bit). The installer is only ~12 MB, requires no administrator privileges (per-user install), and comes as a self-contained native binary with zero external runtime dependencies.

Rust Engine & Flutter UI•Windows 10 & 11 (64-bit)•100% Free & Local

Find and remove duplicate photos — safely.

Name: Duplicate Photo Finder for Windows
Brand: Staksoft
Availability: InStock
Rating: 4.95 (340 reviews)

The duplicate photo finder purpose-built for merged phone backups. Catches exact duplicates, WhatsApp-compressed copies, and resized edits across your folders — keeping the highest-resolution original with guaranteed one-copy protection and full undo.

Download for Windows (64-bit)GitHub Release v1.0.0

✓ ~12 MB Installer✓ No Sign-up or Account✓ Zero Cloud Uploads✓ Recycle Bin Safe

100%

Local & Private

Zero telemetry on your photos

4-Tier

Matching Engine

From SHA-256 to Pixel Correlation

98.6%

Burst Rejection

Never loses distinct memory shots

< 0.5s

Warm Rescan

SQLite cached hash + pixel grids

Duplicate Photos

Windows 11 x64

─□✕

Duplicate Photos

FOLDERS TO SCAN

🗀 E:\dupe-test-corpus×

PROTECTED

Nothing protected. Files here are scanned and shown, but can never be removed.

NEVER SCAN

Nothing excluded. These folders are not read at all.

HOW CLOSELY TO MATCH

Exact

Byte-identical files only. Never a false match.

Strict

Re-saves and metadata edits of the same picture.

Balanced

Recommended. Catches WhatsApp and resized copies.

Loose

Widest net. Review every group before acting.

✓

Include videosMatched by exact contents only

↺ Cleanup historyⓘ About

Files go to the Recycle Bin, never deleted outright. One copy of every photo is always kept.staksoft.com

168.1 MB

could be freed

duplicate groups

extra copies

109

files checked

Scanned in 113 ms

Any size ▾

Any time ▾

Minimum match

Any

Sort Space to free ▾

✓

▶

01 - Copy - Copy - Copy.mp4

IDENTICAL4 copies

88.8 MB

02 - Copy - Copy.mp4

IDENTICAL3 copies

66.7 MB

staksoft logo - Copy.mp4

IDENTICAL2 copies

5.8 MB

IMG_2024009_120000.jpg

SIMILAR5 copies

776.1 KB

IMG_2024024_120000.jpg

SIMILAR3 copies

766.1 KB

4 copies

Byte-for-byte identical

Could free

88.8 MB

📌

▶

BEST COPY01 - Copy - Copy - Copy.mp4

Size: 29.6 MBType: VideoModified: Jun 7, 2026

E:\dupe-test-corpus

✓

▶

01 - Copy - Copy.mp4

Size: 29.6 MBType: VideoModified: Jun 7, 2026Match: Identical

E:\dupe-test-corpus

✓

▶

01 - Copy.mp4

Size: 29.6 MBType: VideoModified: Jun 7, 2026Match: Identical

E:\dupe-test-corpus

✓

▶

01.mp4

Size: 29.6 MBType: VideoModified: Jun 7, 2026Match: Identical

E:\dupe-test-corpus

Why Generic Cleaners Fail

Built for the mess of merged phone backups.

Traditional tools check file hashes. When you back up phones, messengers re-compress every image, turning identical memories into different files on disk.

The 'WhatsApp & Messenger' Duplication Trap

The Danger: When you back up your phone and camera to your PC, the same photo exists in three forms: original 12MB camera shot, 250KB WhatsApp re-compressed copy, and Instagram square crop. Hash-only tools miss them all.

Our Solution: Our Tier-3 (dHash + pHash) & Tier-4 (128x128 Pearson correlation) perceptual engine detects re-compressed and resized copies instantly while preserving the high-res original.

Burst Mode & Rapid-Fire Destruction

The Danger: Standard similarity finders use crude 64-bit perceptual hashes. In burst shots of kids or pets, 8 different photos have the exact same 64-bit hash. Generic tools merge them and delete 7 unique memories.

Our Solution: We test multi-scale resolution divergence (32x32 to 256x256). Re-encodes increase in agreement, while burst frames diverge in detail — rejecting 98.6% of burst false-positives.

Accidental True-Original Deletion

The Danger: Most duplicate cleaners use 'Union-Find chaining' (A matches B, B matches C, so A and C are merged). If B is a middle copy, you can easily end up deleting unique originals.

Our Solution: We enforce strict Star-Topology Clustering: Every duplicate in a group MUST directly match the designated Keeper. A photo that exists in only one place is structurally impossible to remove.

Cloud Ghost Files & Orphaned Live Photos

The Danger: Generic cleaners trigger silent downloads of 50GB from OneDrive / iCloud offline placeholders, or delete an iPhone photo while leaving behind broken `.MOV` Live Photo companion clips.

Our Solution: Full cloud placeholder detection skips offline files, while companion-file awareness links `.MOV`, `.XMP`, and `.AAE` sidecars so they travel safely together.

Under The Hood: Rust Core Architecture

4-Tier Perceptual Matching Engine

Each tier only inspects what previous tiers could not resolve, delivering lightning speed without sacrificing mathematical precision.

Tier 1

SHA-256 Exact Byte Hash

Microseconds

What It Catches

Byte-for-byte identical copies anywhere on your drives.

Technical Implementation

Blazing fast Rust streaming SHA-256 reader with early-exit file size filter.

Precision You Control

4 Tailored Sensitivity Presets

The candidate radii only decide what gets evaluated. The final Tier-4 correlation decides what gets grouped, so wider presets never sacrifice safety.

Exact

Byte-identical

Strict byte-level duplicates only. Zero false positives guaranteed.

dHash Search:-

pHash Search:-

Min Correlation:1.000

Best for: Disk cloning checks, backup folder verification

Strict

Lossless Re-saves

Catches metadata edits, format conversions (PNG to JPG), and re-saves.

dHash Search:6

pHash Search:8

Min Correlation:0.995

Best for: Photography archives, studio asset cleanups

RECOMMENDED

Balanced

Recommended

Catches WhatsApp, Telegram, and messenger re-compressions and resized copies.

dHash Search:10

pHash Search:12

Min Correlation:0.990

Best for: Merged phone backups, family photo libraries, Google Photos dumps

Loose

Manual Review

Widest net for finding cropped, heavily edited, or filtered images.

dHash Search:14

pHash Search:16

Min Correlation:0.970

Best for: Creative asset libraries, visual research archives

Zero-Risk Guarantee

Safe by design. Never lose a photo.

Engineered so that a cleanup can never cost you a memory. Enforced directly in the Rust core, not merely in the UI.

Windows Recycle Bin + 1-Click Undo

Nothing is permanently unlinked. Removals move safely to the Windows Recycle Bin via Win32 Shell API. Every cleanup batch is logged with a single-click instant restore.

Star-Topology Invariant (Always Keeps One)

The Rust engine enforces a strict safety invariant: selecting every item in a duplicate group will NEVER remove the designated Keeper. A photo that exists in one copy cannot be deleted.

Protected & Excluded Folder Boundaries

Mark sensitive archives as 'Protected' (read-only for matching, immune to deletions) or 'Excluded' (completely unread by the scanner).

Live Photo & Companion Sidecar Integrity

Live Photo video clips (.MOV) and RAW sidecar files (.XMP / .AAE) stay coupled with their primary image. If a companion cannot be moved, the photo is held back.

Cloud Hydration Protection

Detects OneDrive, iCloud, and Dropbox offline placeholder files (Reparse Points / Sparse attributes) and skips them, preventing silent multi-gigabyte downloads.

Pre-Flight Modification Verification

Before moving any file, the engine re-checks file size and timestamp against the scan time to ensure sync tools or image editors haven't modified it in the background.

Structural Safety Invariant

Groups are Stars, Not Transitive Chains

Why other cleaners delete unique photos: they merge A→B and B→C into one chain. If B is a middle copy, deleting "duplicates" destroys A.

Other Cleaners (Union-Find)

Burst A ──(close)──> Burst B ──(close)──> WhatsApp C

Transitivity links Burst A and WhatsApp C through B. The app merges all 3 and deletes Burst A — a unique photo that exists nowhere else!

Duplicate Photos (Star-Topology)

Original Keeper <──(confirmed)──> Duplicate Copy

Every candidate MUST directly match the verified keeper. Nothing is merged transitively. A single-copy photo can never be selected or deleted.

More Than Just Duplicates

Find and clear all photo clutter

Surfaces the secondary junk eating up your hard drive — sorted and organized with no automatic risky selections.

Star Grouped

True Duplicates

Exact and perceptual duplicate groups with automatic keeper selection.

Safe Separation

Similar Photos

Different shots of the same scene kept separately so you never lose distinct moments.

Clutter Finder

Screenshots

Surfaces all PNG & JPEG screen grabs across all subfolders in one clean view.

Space Saver

Forwards & Graphics

Identifies memes, stickers, infographics, and low-detail messenger graphics.

Sharpness Rank

Blurry Photos

Ranked least-sharp first using Laplacian variance analysis. Nothing is auto-deleted.

Quick Wins

Largest Space Hogs

Top-down list of heavy RAW files, 4K videos, and massive panoramas.

Simple 4-Step Workflow

Clean your photo mess in minutes

1. Select Your Folders

Pick your Pictures library, external hard drive, or merged phone backup folders. Add protected folders that should stay read-only.

2. Scan with 4-Tier Precision

The Rust engine hashes and matches exact and perceptual copies in seconds, caching results in SQLite for instant future rescans.

3. Review with Smart Select

Examine groups side-by-side. The app auto-selects lower-resolution and compressed copies while preserving the pristine originals.

4. Clean Safely with Full Undo

Send duplicates to the Windows Recycle Bin with one click. Check your storage dashboard to see reclaimed space, and undo anytime.

Real-World Benchmark

7,547 files / 35 GB (Merged phone backup, PC dump, 2,992 WhatsApp images)

Cold Initial Library Scan53.6 seconds

Warm Rescan (SQLite Cache)0.03 seconds (Hash) + 0.30s (Verify)

Burst Rejection Rate98.6% (Zero false burst deletions)

Total Reclaimable Space Recovered283.3 MB of true duplicate waste

Technical Specifications

Native Windows build requirements

Target Operating SystemWindows 10 & Windows 11 (64-bit)

Architecturex86_64 Native (Rust Engine + Flutter UI)

Supported Image FormatsJPEG, PNG, WebP, TIFF, GIF, BMP

Supported Video FormatsMP4, MOV (Exact Byte Matching)

DependenciesNone (Zero .NET, Java, or C++ runtimes required)

Installation TypePer-User (No Administrator Privileges required)

Storage Footprint~12 MB Download / ~31 MB Installed

Data Telemetry0% — 100% Offline, Zero Cloud Uploads

Comparison Matrix

Why Duplicate Photos is in a League of its Own

Feature & Safety Metric	Duplicate Photos (Free)	Generic Cleaners	Manual Sorting
Exact Byte Matching (SHA-256)			Very Slow
Perceptual Match (WhatsApp / Resized / Recompressed)			Nearly Impossible
Burst Shot Protection (Multi-scale correlation)			Subjective
Star Topology (Guaranteed 1 keeper preserved)			High Error Risk
Safe Recycle Bin & 1-Click Undo		Limited / Risky
Live Photo & XMP Companion Awareness			Complex
Cloud Placeholder (OneDrive/iCloud) Safety			Accidental Downloads
Privacy (100% On-Device, No Account)		Often Requires Login
Pricing & Licensing	100% Free	$30 - $60 / year	Hours of Life

Answer Engine Optimized Knowledgebase

Frequently Asked Questions

Deep technical answers explaining matching thresholds, burst protection, and storage safety.

Clean your photo library safely today.

Free, native Windows app. Reclaim gigabytes of wasted storage across merged phone backups with 100% on-device privacy and instant undo.

Download Duplicate Photos (Free)View v1.0.0 on GitHub

Windows 10 / 11 (64-bit)~12 MB InstallerNo Account Required