The JFIF File Format, Explained Byte by Byte

This page is for the people who want the actual bytes. If you only need to open or convert a file, start with what a JFIF file is instead.

The problem JFIF was invented to solve

ITU-T T.81 — the original 1992 JPEG standard — specifies a compression algorithm and an exchange syntax. It does not specify a file format. It says nothing about which colour space the encoded components represent, nothing about pixel aspect ratio, and nothing about physical resolution.

In practice this meant a JPEG stream was not portable: the encoder and decoder had to already agree on conventions the standard never wrote down. JFIF is that agreement, made explicit and machine-readable. Its normative content is small, which is exactly why it succeeded.

Marker structure

A JPEG file is a sequence of markers. Every marker is two bytes: 0xFF followed by a non-zero identifier byte. A JFIF file begins like this:

FF D8                    SOI   Start of Image
FF E0 00 10              APP0  segment, length 16 bytes
4A 46 49 46 00           "JFIF\0"  identifier
01 02                    version 1.02
00                       density units (0 = aspect ratio only)
00 01 00 01              X density = 1, Y density = 1
00 00                    thumbnail width = 0, height = 0
...                      DQT, SOF0, DHT, SOS, entropy-coded data
FF D9                    EOI   End of Image

The FF D8 FF E0 opening is the signature file-identification tools look for. Note that the length field (00 10) counts itself but not the two marker bytes.

The APP0 segment field by field

All multi-byte integers are big-endian. The embedded thumbnail is stored uncompressed, so it is almost always omitted — a 64×64 thumbnail would add 12 KB of raw RGB to the file.

The colour space rule

JFIF's most consequential clause is the one that is easiest to miss: image data must be encoded as either a single greyscale component or three Y′C_BC_R components using the ITU-R BT.601 conversion, with full-range (0–255) values.

That is the reason a JPEG produced by a camera in Tokyo displays with correct colour in a browser in Berlin without any negotiation. Raw T.81 offers no such guarantee.

JFIF versus Exif

Exif (JEITA CP-3451, now ITU-T T.872) is the competing convention, and it is what digital cameras and phones actually write. It uses the APP1 marker instead of APP0 and carries a TIFF-structured metadata block: exposure, aperture, ISO, lens, timestamp, GPS coordinates, orientation.

The two are not mutually exclusive — a file may legally contain both segments, and many do. Strictly, Exif requires APP1 to be the first segment after SOI, so files carrying both are technically non-conformant to Exif, but every real-world decoder tolerates it.

This distinction has a practical consequence worth remembering: files saved as .jfif by a browser usually have no Exif block at all, which means no GPS coordinates and no camera identification. Stripping metadata is a privacy improvement, not a defect.

The JFXX extension segment

JFIF defines a second, optional segment identified by the ASCII string JFXX, also carried in APP0. Its only purpose is to allow a thumbnail stored more efficiently — as a JPEG-compressed image or as a palettised one — rather than as raw RGB. It appears rarely in the wild.

How to identify a JFIF file reliably

Do not trust the extension. Read the first four bytes:

• FF D8 FF E0 with 4A 46 49 46 00 at offset 6 → JFIF
• FF D8 FF E1 with 45 78 69 66 00 at offset 6 → Exif
• FF D8 FF DB → JPEG with a quantisation table first, no APP header
• Anything starting FF D8 FF → some JPEG variant

On Linux or macOS, xxd -l 20 image.jfif shows this immediately. On Windows, Format-Hex -Path image.jfif -Count 20 in PowerShell does the same.

Primary sources

• ITU-T Recommendation T.871 (05/2011), Information technology – Digital compression and coding of continuous-tone still images: JPEG File Interchange Format (JFIF) — itu.int
• ITU-T Recommendation T.81 (09/1992), the base JPEG standard — itu.int
• ITU-R BT.601, studio encoding parameters that define the Y′C_BC_R conversion