⚡Categorize photos & videos in seconds
Whether for digital investigations or for organizing and enhancing personal photo collections, Media Insight scans images and videos, extracts metadata, auto-labels content, lets you run free-text natural language searches, and lets you explore results with blazing-fast search, filters, and similarity lookups.
Media Insight scans and indexes photos and videos, extracts metadata and multimodal embeddings, supports natural-language semantic search, and saves everything into a high-performance local database built to handle millions of items. Reopening projects is instant — no re-processing required.
Download
Choose the build that fits your workflow.
Windows Installer Portable ZIP Mirror Windows Installer
Version 2.6.1 Changelog •
SHA256: 8a3117f4a321cc6906d343ebb695306144bf5bde78a9c4389472c16ad2cc56bc
No cloud. No tracking. Yours forever. 7-day full trial · €10 lifetime license
Preview Video
See Media Insight in action—folder scanning, face detection, EXIF/GPS parsing, and similarity search.
Quick Start
1) Select a folder
Click Select Folder. If the folder was analyzed before, you can load results directly from the database or run a new analysis.
2) Watch the analyzer
Progress shows processed/total, device (CPU/GPU), speed (img/s), and ETA. Video files are sampled into preview frames automatically.
Text Search & SQL Filters
In addition to natural-language search, Media Insight lets you filter images and videos using a simple SQL-like syntax inside the search box. This provides precise control over attributes such as nudity, face count, metadata text, collections, labels, and more.
📌 Basic Field Filters
You can filter using these fields:
face_count > 1restricted = 0nudity > 0known_material = 1selected = 1review_rank >= 2md5 = 'abcdef...'file_name LIKE '%jpg'path LIKE '%DCIM%'exif_full_text LIKE '%iPhone%'
📌 Collections & Labels
Use shortcuts for taxonomy filters:
collection:"Vacations"label:"People"collection:"Beach" AND label:"Family"
📌 Combining Conditions
You may use AND / OR to build more advanced queries:
nudity = 0 AND collection:"Travel"face_count > 0 AND exif_full_text LIKE '%2024%'label:"Portrait" OR label:"Street"
📌 Free-Text Search
If your input is not a SQL expression, Media Insight performs a semantic and text search across:
- file names
- full EXIF text
- folders/paths
- collection names
- label names
Examples: beach sunset, canon lens, family 2023
✔ Examples
collection:"Holidays" AND nudity = 0label:"People" AND face_count > 1exif_full_text LIKE '%iPhone%' AND review_rank >= 2collection:"Beach" OR collection:"Lake"
Key Capabilities
Semantic Search
Search images using natural-language descriptions. Instead of keywords, type free-form text — “a red car at night”, “people walking in a park”, “a dog near the water” — and Media Insight uses semantic embeddings to match scenes by meaning, not just by filename or EXIF.
Face detection & search
Detects faces and generates embeddings. Later, pick any thumbnail and run Find similar faces to retrieve look‑alikes via FAISS.
NSFW & Restricted signals
Classifies nudity (NSFW) and assigns a generic Restricted flag for content that may require special review. You can re‑label items at any time.
Image embeddings
Extracts global image vectors to enable Find similar images across the whole dataset.
Metadata & GPS
Parses EXIF (camera, lens, exposure, etc.) and compact GPS coordinates. Open the spot on Google Maps right from the thumbnail toolbar.
Known‑hash matching
Optional SQLite database of MD5 hashes flags known material (images & original videos). You can create an empty DB from the menu.
Video handling
For each video, Media Insight extracts representative frames (≈10%, 50%, 90%) and indexes them like images. One click exports the original videos, not the frame copies.
Smart caching
Loads previously analyzed folders and similarity data instantly from cache, reducing re-processing time and speeding up navigation.
Collections & Labels
Create custom collections and apply labels for rapid categorization of images and videos. Useful both for investigative casework and for organizing personal media libraries quickly and consistently.
Smart Categorization & Ranking
Automatically assigns categories and applies a ranking system that highlights non-reviewed material first, ensuring the analyst’s attention is focused where it matters most.
Image Analysis (v2.0)
Workspace overview
A dedicated workspace for forensic image inspection. All processing is local; no telemetry.
See live screenshots on the Screenshots page.
Tools included
- Image View — direct rendering for baseline inspection.
- Metadata
- File Header — hex + parsed EXIF
- Metadata Details — group / description / value
- Thumbnail Check — extracted vs difference map
- Social Media
- Social Media Detection — quick markers for platform-specific traces
- Image Analysis
- Double JPEG & Grid
- JPEG Ghost Detection
- Quantization & Huffman Check
- Error Level Map (ELA) with quality/detail controls
- Contrast Adjustment by channel consistency
- Clone Detection (e.g. BRISK) with sensitivity/threshold/grouping
- Filter Detection via texture/probability maps
- Resampling Check with Fourier/probability analysis
- PRNU Fingerprint
- Camera attribution via sensor pattern noise
- Build device fingerprints from image sets
- Export portable
.fprnufiles and reuse them for future comparisons
- Generative Image Detection
- DeepFake Check
- AI-Generated Check
- Categorization
- Manage — manual labeling and quick re-classification
Gallery & Selection
Thumbnails with tools
Each tile shows a toolbar: open file location, see detected faces count, NSFW/Restricted/Known badges, view EXIF, open GPS in Maps, or start face/image similarity.
- Click to toggle selection; Shift‑click to select a range.
- Drag a rectangle on the canvas to multi‑select (toggle mode).
- New icons indicate geographic proximity, same capture date, and identical camera model between items.
Filters
Use the context menu to switch views:
- All, Selected
- Exif, Gps, Faces
- Pornography, Restricted, Known, Safe
Free‑text search matches file names, paths, and EXIF full‑text.
Similarity Search
After an analysis is loaded from the database, the FAISS indices are (re)built in the background. From any thumbnail you can launch Show similar images or, if faces are present, Find similar faces. Results open as a paged view ranked by distance.
- Works with both face and global image embeddings.
- Runs only when the app is idle (not analyzing/loading) to keep the UX responsive.
Database, Paging & Export
Persisted results
Analyses are saved to a local SQLite database. When you re‑open a folder, you can load from DB without re‑processing. Paging keeps large sets smooth; a status line confirms the current page and total.
Export selected
Export copies of selected images and the original videos (even if you selected extracted frames). Media Insight creates a tidy MediaInsight_Export folder with sub‑folders for images/videos.
Managing Videos & Storage
You can bulk‑clean the auto‑created im-video-frames folders across all analyzed directories when you need to reclaim disk space.
Labels & Re‑classification
You can manually set an item to Pornography, Restricted, or Safe from the menu. This immediately updates both the UI and the database.
Portable Installation
When Media Insight is installed on an external USB drive, no data is written to the target device during a search and seizure of electronic devices/data, ensuring a clean and portable workflow for forensic use.
Licensing
Media Insight checks for a valid license (or remaining trial) when you open a folder. If needed, a login dialog appears to activate access; background refresh keeps the session valid.
Notes on the term “Restricted”
Restricted is a generic label meaning “content that may require special review or limited access.” It intentionally avoids naming any specific category so the software remains suitable for general‑purpose workflows.