Extract Selected Glyphs from a Linux Console Font with splitfont

splitfont is a tiny, old kbd utility that pulls individual glyphs out of a raw Linux console font, one file per character. It comes from kbd 2.6.4-2ubuntu2 here. Give it about fifteen minutes.

• You need: the kbd package, and a font in the exact format this old tool expects: no PSF header, exactly 256 characters.
• It will not: load a font, touch the console, or need sudo. Work on a temporary copy.

1. Check the installed command

Confirm the binary and its short usage message:

$ command -v splitfont
/usr/bin/splitfont
$ dpkg-query -W -f='${Package} ${Version}\n' kbd
kbd 2.6.4-2ubuntu2
$ splitfont
call: splitfont fontfile 17,23-30,...

The syntax is splitfont FONTFILE RANGE..., where a range is a comma-separated list of decimal character numbers and inclusive ranges, such as 17,23-30. The manual describes the expected input as an ISO-type font with 256 characters and no PSF header.

2. Prepare a safe working copy

Never experiment on the only copy of a font. This example takes the 4,096 glyph bytes from an installed 8 by 16, 256-character PSF font, after its 32-byte PSF2 header. It is a test fixture for splitfont, not a general PSF conversion recipe:

$ workdir=$(mktemp -d /tmp/splitfont.XXXXXX)
$ dd if=/usr/share/consolefonts/UbuntuMono-R-8x16.psf \
    of="$workdir/font.raw" bs=1 skip=32 count=4096 status=none
$ stat -c '%n: %s bytes' "$workdir/font.raw"
/tmp/splitfont.XXXXXX/font.raw: 4096 bytes

A 256-character font at 16 bytes per glyph comes to 4,096 bytes. Your font may use a different glyph height, so calculate the real raw size rather than copying these numbers blindly. Already have a headerless font? Copy it to a temporary directory under a new name and work there instead.

Skip sudo for this step unless the source is unreadable to your account. If you genuinely need elevated access, copy the source to a directory you own first, then run splitfont on that copy.

3. Extract individual characters

Pick character numbers and run the command from the directory holding the working copy:

$ cd "$workdir"
$ splitfont font.raw 65,66-67
$ printf 'status: %s\n' "$?"
status: 0
$ ls -l font.raw.41 font.raw.42 font.raw.43
-rw-r--r-- 1 you you 16 ... font.raw.41
-rw-r--r-- 1 you you 16 ... font.raw.42
-rw-r--r-- 1 you you 16 ... font.raw.43

• Decimal in, hex out. 65, 66 and 67 are the requested decimal character numbers, but the generated suffixes are hexadecimal: .41, .42, .43.
• One glyph per file. Each output is exactly the font's glyph height, 16 bytes here.

Ranges are inclusive, so 23-30 asks for eight consecutive characters. Keep the range list explicit in scripts, and quote the font path if it contains spaces:

$ splitfont '/path/to/raw font' 32,48-57,65-90
$ printf 'status: %s\n' "$?"
status: 0

4. Verify the result and the source

Check the extracted files have the right size, and that the source came through unaltered:

$ sha256sum font.raw > before.sha256
$ splitfont font.raw 65
$ sha256sum -c before.sha256
font.raw: OK
$ stat -c '%n: %s bytes' font.raw.41
font.raw.41: 16 bytes

splitfont writes extracted files beside the input and, in this test, does not touch the input itself. Treat that as an observed property of this installed build, not a licence to skip backups when the font actually matters. If an output file with the same name already exists, stop and inspect it first; the command may silently overwrite it.

To compare a glyph against its position in the raw source, use a byte offset of the character number times the glyph height. For character 65 in this 16-byte font:

$ dd if=font.raw bs=16 skip=65 count=1 status=none | cmp - font.raw.41
$ printf 'compare status: %s\n' "$?"
compare status: 0

5. Handle the common input errors

Seeing file unexpectedly large means the input is not the required 256-glyph, headerless size. Check its size and format before touching anything:

$ file /path/to/font
$ stat -c '%s bytes' /path/to/font

The installed program also rejects a file whose size is not a multiple of 256, but that check will not spot a PSF header for you, and a file that passes it can still hold the wrong data. Use a known font format and a deliberate conversion step, not a guess.

A mistyped range returns a non-zero status and a bad-argument message. Do not infer the output name from a decimal suffix: the actual names are hexadecimal. If the command cannot open the file, fix the path or read permission rather than reaching for root to paper over the mistake.