Python API

Audio

The Python-visible decoder, encoder, feature extraction, and signal-processing surface.

Generated from Source/Python15 functions9 classes1 enums

This inventory is generated from symbols registered by OA's nanobind source and public Python facade. It deliberately excludes unbound C++ internals. Signatures show the Python keyword names declared by each binding; follow the source link for ownership, capability, and execution details.

oa.audio

Functions

AmplitudeToDb(buffer, floor_db)

20·log10(max(|x|, floor)) — amplitude to dB with a silence floor.

Clip(buffer, min, max)

Hard clip to [min, max].

Fade(buffer, fade_in_samples, fade_out_samples)

Linear fade-in/out envelope (sample counts per edge).

Gain(buffer, gain_db)

Scalar gain in dB.

MelSpectrogram(buffer, sample_rate, config)

Mel spectrogram → [Channels, NumMels, Frames] F32 (Whisper/CLAP layout).

Mfcc(buffer, sample_rate, config)

MFCC → [Channels, NumCoeffs, Frames] F32 (orthonormal DCT-II of log-mel).

Mix(a, b, gain_a, gain_b)

Weighted sum gain_a·a + gain_b·b.

Normalize(buffer, target_db, mode)

Peak (mode 0) or RMS (mode 1) normalization to target_db.

OaChannelsForLayout(layout)

Expected channel count for a layout (0 = unknown).

OaLayoutForChannels(channels)

Best-effort layout for a raw channel count.

PreEmphasis(buffer, alpha)

Pre-emphasis filter y[n] = x[n] − α·x[n−1] (y[0] = x[0]).

Resample(buffer, in_rate, out_rate, filter_half_width)

Windowed-sinc resample → [Channels, Samples·out/in] F32.

Stft(buffer, config)

STFT magnitude → [Channels, Frames, FftSize/2+1] F32.

ToMatrix(buffer)

Zero-copy reshape of raw waveform to [Channels, 1, SampleCount].

ToMono(buffer)

Average channels → [1, Samples].

Classes

OaAudioDecoder2 members

LoadFile(path)

Decode a WAV/FLAC/MP3 file to a planar [Channels, Samples] F32 GPU buffer.

static

LoadMemory(data)

Decode audio from an in-memory byte buffer (bytes; WAV/FLAC/MP3).

static

OaAudioDecodeResult7 members

Buffer(...)

Decoded planar [Channels, Samples] F32 GPU buffer.

prop_ro

SampleRate(...)

ro

ChannelCount(...)

ro

SampleCount(...)

ro

IsValid(...)

method

DurationSeconds(...)

method

Meta(...)

Metadata POD for the OaFnAudio ops (layout inferred from channels).

method

OaAudioEncoder2 members

EncodeWavF32(buffer, sample_rate)

Encode a [Channels, Samples] F32 GPU buffer to WAV-F32 bytes (synchronous).

static

SaveWavF32(path, buffer, sample_rate)

Encode + write a lossless WAV-F32 file (synchronous file/codec boundary).

static

OaAudioMeta6 members

__init__(...)

constructor

SampleRate(...)

rw

ChannelCount(...)

rw

SampleCount(...)

rw

Layout(...)

rw

DurationSeconds(...)

Clip duration in seconds (SampleCount / SampleRate).

method

OaMelConfig8 members

__init__(...)

constructor

FftSize(...)

rw

HopSize(...)

rw

NumMels(...)

rw

FMin(...)

rw

FMax(...)

Highest frequency in Hz (0 = SampleRate/2).

rw

LogScale(...)

rw

Normalize(...)

Per-channel instance normalization.

rw

OaMfccConfig3 members

__init__(...)

constructor

NumCoeffs(...)

rw

Mel(...)

rw

OaNormalizeAudioConfig3 members

__init__(...)

constructor

Mode(...)

0=peak (max abs), 1=RMS

rw

TargetDb(...)

rw

OaResampleConfig4 members

__init__(...)

constructor

InRate(...)

rw

OutRate(...)

rw

FilterHalfWidth(...)

rw

OaStftConfig6 members

__init__(...)

constructor

FftSize(...)

rw

HopSize(...)

rw

WinSize(...)

rw

Window(...)

0=Hann, 1=Hamming, 2=Blackman, 3=Rect

rw

Center(...)

rw

Enums

OaChannelLayoutSource

Mono · Stereo · Surround5_1 · Surround7_1 · Unknown