About

MP4 files conform to ISO 14496-12 (ISO Base Media File Format). The container organizes data into nested boxes (atoms), each identified by a 4-character code and a 32-bit or 64-bit size field. Critical metadata lives in the moov atom: the mvhd box stores global timescale and duration, while each trak contains per-track timing (mdhd), handler type (hdlr), and codec parameters (stsd). Timestamps in MP4 use an epoch of January 1, 1904 - not Unix 1970 - meaning raw values require an offset of 2082844800 seconds. Getting codec details wrong leads to transcoding failures, streaming incompatibility, or silent audio.

This tool parses the binary structure directly in your browser using DataView on the raw ArrayBuffer. No file leaves your machine. It reads fixed-point fields (16.16 for rates, 2.30 for volume), decodes iTunes-style ilst metadata (artist, title, album art), and maps hdlr component subtypes to human-readable track labels.

Limitation: heavily fragmented files with moof atoms (live streams) may show incomplete track data since the tool targets standard progressive MP4 layout.

Formulas

Duration in seconds from raw MP4 fields:

T = durationtimescale

where duration is the tick count from mvhd or mdhd, and timescale is ticks per second (commonly 600 for video, 44100 or 48000 for audio).

MP4 timestamp to Unix timestamp conversion:

t_unix = t_mp4 − 2082844800

where 2082844800 is the offset in seconds between 1904-01-01 and 1970-01-01.

Fixed-point 16.16 decoding:

value = raw₃₂65536

Average bitrate estimation:

bitrate = fileSize × 8T bps

where fileSize is in bytes and T is duration in seconds.

Reference Data

Atom / Box	ISO Code	Contains	Typical Location
File Type	ftyp	Brand, version, compatible brands	Root (first box)
Movie Header	mvhd	Timescale, duration, creation/modification dates, rate, volume	moov
Track Header	tkhd	Track ID, dimensions, transformation matrix, duration	moov → trak
Media Header	mdhd	Track timescale, duration, language code	moov → trak → mdia
Handler Reference	hdlr	Track type: vide, soun, text, sbtl	moov → trak → mdia
Sample Description	stsd	Codec FourCC (avc1, hev1, mp4a, etc.), sample rate, channels	moov → trak → mdia → minf → stbl
Sample Table	stbl	Sample sizes, chunk offsets, sync samples	moov → trak → mdia → minf
User Data	udta	Custom metadata, copyright, location	moov or moov → trak
iTunes Metadata	ilst	Artist, title, album, cover art, genre, year	moov → udta → meta
Free Space	free	Padding bytes (ignored by players)	Root or moov
Media Data	mdat	Actual audio/video sample data	Root
Edit List	elst	Media time offsets, dwell segments	moov → trak → edts
Composition Offset	ctts	PTS/DTS offset for B-frames	stbl
Sync Sample	stss	Keyframe indices	stbl
Color Parameter	colr	Color primaries, transfer characteristics, matrix	stsd entry
Pixel Aspect Ratio	pasp	hSpacing, vSpacing	stsd entry
AVC Config	avcC	Profile, level, SPS/PPS NAL units	stsd → avc1
HEVC Config	hvcC	Profile, tier, level, VPS/SPS/PPS	stsd → hev1/hvc1
Audio Config	esds	AAC profile, sample rate, channels	stsd → mp4a
GPS Location	©xyz	Latitude, longitude (ISO 6709)	udta

Frequently Asked Questions

MP4 files use an epoch starting January 1, 1904, per ISO 14496-12. A raw value of 0 maps to that date. Some encoding tools write 0 when the actual timestamp is unknown. The viewer applies the 2082844800-second offset to convert to a standard date, but if the source wrote garbage data, the output will be unreliable.

The mvhd timescale is a global movie timescale (often 600 or 1000). Each track has its own mdhd timescale matching its media clock - typically 90000 for H.264 video, 44100 or 48000 for AAC audio. Duration in either box must be divided by its respective timescale to get seconds.

Yes, if the recording device (phone, drone, GoPro) wrote a ©xyz atom inside udta. The value follows ISO 6709 format: a string like +34.0522-118.2437/ representing latitude and longitude in decimal degrees. Not all MP4 files contain this data.

The tkhd box stores the track's presentation dimensions after the transformation matrix is applied. The actual coded resolution sits in the stsd sample entry (e.g., avc1 width/height). If a pasp (pixel aspect ratio) box is present, the display width is adjusted by the ratio hSpacing/vSpacing. This viewer shows both coded and display dimensions.

The viewer calculates average bitrate as file size in bits divided by duration in seconds. This is an overall container-level estimate including all tracks. True per-track bitrate requires parsing every sample size from stsz, which this tool summarizes as sample count. For VBR content, instantaneous bitrate varies significantly from this average.

Fragmented MP4s used in DASH/HLS streaming store samples in moof+mdat pairs rather than a single moov+mdat structure. This tool parses the moov/moof boxes it finds, but track duration and sample counts may be incomplete since each fragment is self-contained. For full analysis of fMP4, concatenate all segments first.