Module mpeg 

MP3 stream info the way mutagen computes it. A transcription of mutagen.mp3.MPEGInfo and mutagen.mp3._util.

The Python pipeline reads durations and bit rates from mutagen headers, not from a decode, and mutagen disagrees with ffprobe on roughly one file in six of the real library (encoder delay handling, LAME padding, CBR estimates). A report that says 2.4s where Python says 2.5s fails parity, so the arithmetic here follows mutagen line for line, including its file-size fallback and its half-to-even bit rate rounding.

Structs

Frame 🔒

One parsed frame header plus whatever the Xing/VBRI/LAME tags in it said. length is None where mutagen leaves the attribute unset.

MpegInfo

MPEGInfo’s useful attributes.

Xing 🔒

Enums

MpegError

What mutagen raises, by the name Python code prints via type(exc).__name__ — that name reaches CSV output in sfx-impact.

Version 🔒

Functions

bitrate_table 🔒

Per-(version, layer) bit rate tables in kbps, index by the 4-bit field.

clone_frame 🔒

duration_ms

_mp3_duration_ms: int(info.length * 1000).

info

MP3(path).info.

info_from

MPEGInfo(fileobj): sync to the first believable frame run and derive length and bit rate from it.

intround 🔒

intround: Decimal.from_float(v).to_integral_value(ROUND_HALF_EVEN).

parse_frame 🔒

Parse the frame at pos. None is mutagen’s HeaderNotFoundError.

parse_lame 🔒

LAMEHeader.parse_version + the two fields of LAMEHeader the length needs. None wherever mutagen raises LAMEError or reports no extended header.

parse_vbr_header 🔒

A Xing/Info tag trumps the header’s bit rate and gives a real length; failing that, a Fraunhofer VBRI tag. Neither raises.

parse_vbri 🔒

VBRIHeader: (bytes, frames) when a valid tag sits at pos.

parse_xing 🔒

read_exact_at 🔒

sample_rates 🔒

skip_id3 🔒

skip_id3: step over every leading ID3 tag (WMP stacks them).

u16be 🔒

u32be 🔒