Module audioop 

The audioop primitives pydub is built on, transcribed from CPython’s Modules/audioop.c (shipped for Python 3.13+ as audioop-lts).

Every conversion keeps the C code’s arithmetic: products are taken in double, clamped with fbound and then floored (not truncated), so -3 * 0.5 is -2. ratecv interpolates on 32-bit-shifted samples and casts with C’s truncating (int). Those two rounding rules are the difference between a byte-identical DAW layer and one that is merely close, so nothing here is “tidied”.

Buffers are raw little-endian bytes of width 1, 2 or 4, as in C.

Functions

add

audioop.add(fragment1, fragment2, width) — saturating sum.

add_into

In-place add of b onto a[offset..]. Same arithmetic; used where the pydub code builds a new buffer that would be byte-identical anyway.

bounds 🔒

Smallest and largest sample for a width, as doubles like the C tables.

check_len 🔒

fbound 🔒

fbound: clamp into range, with C’s val < minval + 1 lower test.

gcd 🔒

get_raw

GETRAWSAMPLE — the signed sample at byte offset pos.

get_sample32 🔒

GETSAMPLE32 — a sample scaled to occupy the top of an int.

int_bounds 🔒

lin2lin

audioop.lin2lin(fragment, width, newwidth).

mul

audioop.mul(fragment, width, factor).

put_raw

SETRAWSAMPLE — append one sample, truncating to the width.

put_sample32 🔒

SETSAMPLE32 — the inverse, with an arithmetic shift.

ratecv

audioop.ratecv(fragment, width, nchannels, inrate, outrate, None) with the default weights (1, 0). Only the converted bytes are returned; pydub discards the state.

tomono

audioop.tomono(fragment, width, lfactor, rfactor).

tostereo

audioop.tostereo(fragment, width, lfactor, rfactor).