Understand your media.
Find what changed.

Use local AI to find people and vehicles, then review their tracks. Locate text and clock watermarks with OCR. Frame-number seeking, independent Quick View windows and parameter comparison take you back to the relevant frame.

macOS 14+ · Apple Silicon · Local test build

  • Native macOS
  • Local AI & OCR
  • 中文 / English
Explore the workspace
The current MediaMeta workspace: library, player and persistent AI tools, with parameters, checks, and playback advice on the right.
Library, player, analysis tools, and inspector. One file, one continuous context.
Understand parametersCompare parametersAI & OCRCheck & export

Let AI find the subject.
Go straight to the frame.

Detect and track 80 classes of objects, including people and vehicles, on your Mac. Overlays, class timelines and a track list show the first and last observed moments. Click a track to return to a matching frame.

AI detection results: English person and bus labels with confidence scores on the frame, plus English category timelines showing matching frames.
Detection boxes and category timelines stay linked. Select a match to return to its frame.

From detecting a subject to finding a frame you can review.

  • Find observed moments by trackReview each track’s class, ID, first and last matches, and observation count. Click a track to seek to its first match under the current filters, or review matches by class.
  • Focus on an areaDraw a rectangular or polygonal ROI and adjust confidence thresholds by class. Overlays, timelines, tracks and statistics update together without reassigning track IDs when filters change.
  • Export the evidence with annotationsChoose a range and export annotated frames or video, with optional track IDs. Exports retain the filters selected when they start, so you can review the evidence before saving it.

YOLO26 Nano / RF-DETR Nano · Local Core ML inference · No cloud account

  1. Choose a range and model
  2. Find frames by category
  3. Review and export

Follow text and time watermarks along the timeline, too.

OCR extracts on-screen text and locates its moments. Select a clock watermark region to compare the displayed clock with elapsed video time and inspect frames around suspected jumps. Markers and waveforms keep your review in context.

Explore AI & OCR

Detection and tracking help locate and review footage; they do not establish identity, unique person counts or links across videos. Track IDs apply only within one analysis. Sampling can miss brief appearances, and unanalyzed ranges do not establish absence.

From parameters to picture,
in one workspace.

Start with the information you need. Reach for raw fields, track mapping, and advanced controls when the task calls for them.

Understand the file, not just its fields.

Read common parameters first, then explore container, video, audio, and subtitle tracks. Search fields, read explanations, or open the raw JSON.

Learn about parameters

See what changed between versions.

Compare media parameters across multiple files against a baseline. Normalize codecs and units, adjust numeric tolerances and ignored fields, and resolve ambiguous track matches manually.

Explore both kinds of comparison

Check a batch against your requirements.

Use a shared parameter checklist or draft a template from a reference video. Separate mismatches, review items, and unknowns; expand passing items on demand. Parameter compliance is not a complete quality assessment.

Build a checklist from a reference

Review in separate windows.

Each tab is an independent workspace. Open several files in Quick View, each with its own playback state and Dock preview. Seek by frame number in the workbench or Quick View; temporary files enter your library only when you explicitly import them.

Explore workspaces and Quick View

This time’s reference.
Next time’s checklist.

In the parameter page’s More menu, choose “Use as check reference…” to draft editable rules from a parsed video.

Select the codecs, dimensions, frame rate, or audio parameters that must match. Review values, tolerances, and track matching, then save for single-file or batch checks.

Read the template guide
reference.mp4Reference → Review draft → Save checklist

Delivery parameter checklist

Video codec
AVC (H.264)
Dimensions
1920 × 1080
Frame rate
25 fps ± 0.001
Audio sample rate
48,000 Hz

Workflow illustration, not an app screenshot. Available fields depend on parsing; a template is not quality certification.

Know what was checked.
Export that evidence.

Technical checks, picture and sound checks, and parameter checklists have separate records. Export reports that have actually run, with their range, tracks, rules, and unknowns intact.

Choose the right report

Technical checks

Inspect container structure, indexing, timelines, or decoding on demand. The required ffprobe / ffmpeg tools are bundled.

Picture & sound checks

Compare content evidence for selected ranges and tracks. Matching samples do not prove whole-file identity, visual quality, listening quality, or synchronization.

Parameter checklists & automation

Read reports offline, or connect CLI, machine-readable JSON, JUnit, and CI commands to a developer workflow.

From source to delivery,
with fewer detours.

The complete guide
  1. 1

    Add & organize

    Drop in files or folders, then search, filter, and select. Pause, cancel, or retry the parsing queue.

  2. 2

    Compare & locate

    Compare parameters and locate issues with frame numbers, object tracks, OCR or markers. Review files in independent Quick View windows; choose a range before running the checks you need.

  3. 3

    Review & export

    Review against a checklist and inspect privacy fields before sharing. Take a report, frame, or clip—not an unqualified “pass.”

On-device media. Transparent data flow.

Parsing, content analysis, and local AI do not require media uploads. Link checks request the address you specify. App diagnostics are enabled by default when configured and can be disabled in Settings; feedback separately asks you to confirm what is sent.

Pre-sharing checks read parsed privacy fields and additional tracks. They do not scan faces, voices, or filenames, or remove information automatically.

Read privacy & limitations

Before you start,
a few things to know.

What it can do—and what it cannot prove.

Does a passing check mean the whole file is problem-free?

No. Parameter checklists evaluate specified fields. Content checks cover selected ranges, tracks, and modes; technical checks record what actually ran. Playback, completed decoding, and matching parameters are different kinds of evidence.

Do AI and OCR upload my media?

Local AI and OCR process media on your Mac without requiring media uploads. Results help you locate and review content, not certify authenticity or overall quality. Diagnostics and explicit link requests are separate data flows described in the privacy guide.

Do I need additional tools?

The current test build includes MediaInfoLib, ffprobe / ffmpeg, and AI / OCR models. No extra Python or media-tool installation is needed. Check tool status in Settings; local recognition needs no first-run model download.

Does the pre-sharing check remove private information?

No. It inspects parsed metadata and track structure without modifying media or generating an anonymous copy. No findings do not mean the file is anonymous; review picture, sound, filenames, and reports yourself.

Which Macs can run the test build?

The test build is intended for Macs with Apple silicon (M-series chips) running macOS 14 or later. Intel Macs are not supported. Read the test-build installation notes before installing.

Bring your media.
Take the evidence with you.

Start with one file. Explore the tools your task needs.

Public downloads coming soon

0.2.0 · Build 82 · Apple Silicon · macOS 14+This test build is not notarized by Apple. macOS may prevent it from opening.

The current test version is 0.2.0 / Build 82. Public downloads are not available yet. If macOS shows a security warning, read the installation notes first; do not disable security protections.

MediaMeta app icon