Built on advanced on-device AI and hardware acceleration to deliver professional-grade performance with uncompromising privacy.On-device AI — professional-grade, fully private.
Video Metadata Online
Video metadata read on your device — codecs, tags, streams
- Nothing uploaded
- Reads AVI, WMV and FLV too
- Edit tags without re-encoding
Drop a video here
One video at a time, any size. MP4, MOV, MKV, WebM, AVI, WMV, FLV, OGV, 3GP, MPEG and GIF.
Reads the container's index. Instant, whatever the file weighs.
Reads the whole file, so it is the one option whose cost depends on size. Useful for proving two copies are identical.
Everything adds raw container tags and per-track internals.
The complete report, for a script or another tool to read.
Fields you leave alone stay as they are.
Rewrites the container's tag area only. Every video and audio packet is copied untouched and checked afterwards.
Add a video to read its metadata.
What this reader shows you
A video file is a container full of labelled boxes, and almost everything you might want to know about a recording is written in them before a single frame is decoded. This tool opens those boxes and prints what it finds, in full, on your device. Nothing is uploaded and nothing is estimated where the file states a real value.
Every field, grouped the way you would ask for it
- The container: what format it really is, which brands it declares, how long it is, its overall bitrate, how many streams it holds, when it was created, and which software wrote it.
- The picture: codec, profile, level and the exact codec string, the size it is stored at and the size it is meant to be shown at, pixel aspect ratio, rotation, frame rate and whether that rate is constant, the number of frames, key frames and average distance between them, bit depth and chroma subsampling.
- The colour: primaries, transfer curve, matrix and range, with HDR10 and HLG identified by name, plus the mastering display and content light levels when a file carries them.
- The sound: codec and profile, sample rate, channel count and layout, bit depth and bitrate, for every audio track rather than only the first.
- Everything else: subtitle, timecode and data tracks, container tags both normalised and raw, cover art, and attached files such as the fonts an MKV carries for its subtitles.
Every number says where it came from
A bitrate a file declares and a bitrate worked out from its size are different claims with different accuracy, and a report that shows them under the same label is hiding something. Each value here is labelled with its source — declared in the file, read from the sample table, or measured across the whole file — so you can tell what is fact and what is arithmetic.
It opens the containers everything else refuses
Our other video tools accept MP4, MOV, MKV and WebM, and turn everything else away at the door. That is the right behaviour for them: they have to decode frames, and no browser ships a decoder for the codecs an AVI, a WMV, an FLV, an OGV, a 3GP or an MPEG usually carries. This tool is the exception, because reading a header is not decoding.
What each container gives you here
- MP4, M4V, MOV, 3GP and 3G2: the full report, down to the decoder configuration record — profile, level, chroma and bit depth — plus edit lists, rotation matrices, and both iTunes-style and QuickTime-style tags.
- MKV and WebM: the full report from the segment information and track entries, including the colour element, projection rotation, attachments and every tag targeted at the file as a whole.
- AVI: stream headers, the bitmap and wave format structures, OpenDML frame counts for files past two gigabytes, and the RIFF INFO chunk.
- WMV and ASF: file properties, per-stream video and audio format blocks, content description and the extended content descriptors where Windows keeps most of its tags.
- FLV: the onMetaData script tag, which is where duration, dimensions, frame rate and data rates live.
- OGV and Ogg: the Theora, Vorbis, Opus and FLAC identification headers, and the Vorbis comment block.
- MPEG program and transport streams: the sequence header for picture size and frame rate, the program map for what tracks exist, and the audio frame header for sample rate and channels.
- Animated GIF: logical size, frame count, per-frame delays summed into a real duration, loop count, transparency and palette size.
Where a file cannot be played or processed on the web, the report says so plainly and says why, rather than leaving you to find out by trying. If you need to work with one of these files, convert it first with the video converter and the rest of the toolkit opens up.
Will this file play on your device?
This is the question a metadata tool is really being asked most of the time, and it is the one thing a website that reads your file on a server cannot answer — it does not know what your browser is. This one is running on your machine, so it can simply ask.
How the verdict is reached
- The container is checked first. An AVI holding H.264 still will not open in a browser, and blaming the codec would send you off to re-encode a file that only needed remuxing.
- Then each track's codec. The browser is asked directly whether it has a decoder for that exact codec string, at that exact size. Nothing is decoded to find out — configuring a decoder and never feeding it is a question, not work.
- Picture and sound are answered separately. A file whose audio codec your browser cannot decode still plays; it plays silently. That is a useful thing to know and a misleading thing to fold into a single verdict.
- The answer can be “not known”. Where a container does not state its codec precisely enough to ask, the report says so rather than guessing. An unknown reported as a yes is how a file that cannot play gets called playable.
Reading the numbers
Several fields in the report answer questions people arrive with, and are worth a sentence each.
The ones that come up most
- Stored size against displayed size. A phone records in landscape and writes a rotation flag telling players to turn the picture. The stored size is what the encoder produced; the displayed size is what you should see. When software ignores the flag, video appears sideways — and this is where you find out that is what happened.
- Frame rate mode. Constant means every frame lasts the same time. Variable means they do not, which is normal for screen recordings and phone footage and awkward in an editing timeline. The header can suggest it; the full scan settles it.
- Average GOP. How many frames sit between key frames. A large number compresses better and makes seeking and cutting less precise, which is why a file cut on a non-key frame can start with a smear.
- Chroma subsampling. 4:2:0 stores colour at a quarter of the picture's resolution, which is invisible on camera footage and very visible on screen recordings of text. 4:4:4 keeps it all.
- Transfer curve and HDR. The curve, not the colour gamut, is what makes a file HDR. Wide primaries with an ordinary curve is a common and perfectly normal combination, and calling it HDR would be wrong.
- Starts before fully downloaded. An MP4 keeps its index either before or after the video data. Before, and it starts playing as it downloads; after, and a browser has to fetch the whole file first. This is the single most common reason a web video seems slow to start.
Nothing is uploaded, and that is why it is fast
Every other way of doing this sends your file somewhere. That is a privacy question for a holiday video and a serious one for anything under an agreement — and it is also why those tools have a size limit, because someone has to pay for the transfer and storage.
How the reading actually works
A container writes its index in one place, and states the length of everything else so a reader can skip past it. So the tool reads a few small slices of the file and jumps over the video data entirely. Measured on a real MP4 padded out to test it, a 64 MB file, a 256 MB file and a 1 GB file each cost the same 131,072 bytes read and finished in a fraction of a second. There is no size limit on reading because there is nothing to limit.
The one operation that reads everything
Computing a SHA-256 has to look at every byte, so it is off by default and clearly labelled. It runs on your device too, using your processor's own hashing instructions, and it is the fastest honest way to prove two copies of a file are identical.
Reading runs entirely on your device. A report costs a few kilobytes of reading whatever the file weighs, so there is no size limit and nothing to wait for.
How to use it
Five steps, and the first three happen on their own.
Step by step
- Add a video. Drag it onto the page or click Browse. It is read where it sits — nothing is copied or sent.
- Read the report. The container, every stream, the colour information and the tags appear immediately, whatever the file weighs.
- Turn on the full scan if you need certainty. It walks the packet index to measure the real bitrate and to settle constant against variable frame rate.
- Edit or remove the tags. Change what you need and save, or clear everything in one action. The picture and sound are copied untouched and checked afterwards.
- Export. JSON for a script, CSV for a spreadsheet, plain text for a bug report — or copy any of them straight to the clipboard.
When you need this
The report answers a specific question most of the time, and these are the questions it gets asked.
Common reasons people open it
- A file will not upload or play somewhere. Platforms reject files for a codec, a profile, a frame rate or a colour space rather than for their size, and the rejection rarely says which. The report shows all four.
- Something is sideways. Rotation is a flag, not a change to the picture. Seeing the flag tells you whether to rotate the file or to change the software reading it.
- Two files are supposed to be the same. Compare the reports field by field, or compare one SHA-256 line, instead of comparing file sizes and hoping.
- You are about to edit or grade. Frame rate mode, chroma subsampling, bit depth and colour range all decide how a clip behaves on a timeline, and all of them are decided before you import it.
- You are sharing a file publicly. See what the tags say about you — the device, the software, the location fields, the account name an export tool wrote in — then remove all of it in one action.
- You are filing a bug or a support ticket. A text report describes a file precisely in a way that “a 4K MP4 from my phone” does not.
Frequently Asked Questions
What metadata does this tool show?
Everything the file states about itself. The container and its brands, duration, overall bitrate and stream count; per track, the codec, profile, level, coded and display size, pixel aspect ratio, rotation, frame rate and frame rate mode, key-frame count, average GOP, bit depth and chroma subsampling; the colour primaries, transfer curve, matrix and range, with HDR10 and HLG identified; sample rate, channel count, channel layout and bit depth for sound; and every container tag, cover image and attachment, both normalised and raw.
Can I edit metadata with this tool?
Yes. You can change the title, artist, album, album artist, genre, date, comment and description in MP4, MOV, MKV and WebM files, or remove every tag in one action. The video and audio packets are copied byte for byte and only the tag area is rewritten, so nothing is re-encoded and the picture and sound are unchanged. The tool verifies that packet by packet before it offers you the file.
Why can this open an AVI or a WMV when your other video tools cannot?
Because reading a header is not decoding. A container's structure is plain bytes with lengths in them, and understanding those bytes does not require a decoder for what is inside. Our editing and conversion tools need to decode frames, which is why they refuse AVI, WMV, FLV, OGV, 3GP and MPEG — no browser ships a decoder for the codecs those usually carry. This tool reads all of them, and tells you plainly that they cannot be played or processed on the web.
How big a file can I read?
Any size. A report is a header read: measured on a real MP4 padded out to test it, a 64 MB file, a 256 MB file and a 1 GB file each cost 131,072 bytes read and finished in a fraction of a second, because the work does not depend on how long the video is. Editing tags is different — that copies the whole file — so it is capped at 300 MB on the free plan.
Why might the duration here differ from what my media player reports?
A container states a duration in its header, and a player often measures a different one by reading to the end. They disagree when the header was written before the recording finished, when an edit list shifts the timeline, or when the file was cut without rewriting the index. Turn on the full scan and the tool measures the duration the same way a player does, so you can see both numbers and know which your software will use.
What does “variable frame rate” mean here, and how do you know?
It means the gaps between frames are not all the same, which is normal for screen recordings and phone footage and awkward for editing. The tool reads the file's own time-to-sample table, so a single entry means every frame has the same duration. The full scan goes further and examines the real frame timestamps across the whole file, which is the only way to be certain.
Why does the preview sometimes not appear?
Because your browser has no decoder for that codec, or cannot open that container at all. The report says which, in plain words, and it still reads every field either way — the metadata does not depend on being able to play the file. That verdict is measured against your actual browser, not guessed from the file extension.
Is anything uploaded?
No. The file is read on your device and never leaves it. There is no server to send it to: reading, editing and exporting all run locally, which is also why a 1 GB file costs nothing to open.
What is the SHA-256 for?
Proving that the file you are looking at is the file you were sent. Two people can each open their copy here and compare one line of text rather than trusting file sizes and names. It is optional because it is the only operation that reads the whole file.
Does removing metadata make a file anonymous?
It removes the container's tag area — titles, comments, cover art, the encoder's name, the software that made it, and location fields written there. It does not touch anything encoded inside the picture or sound themselves, such as a visible timestamp burnt into the frame or a watermark. For most files the tag area is where the identifying information lives, and the report shows you exactly what was there before you remove it.
from 423 ratings
Rate this tool
Tap a star — it takes a second