Configuration¶

Configuration is represented by immutable dataclasses in Python and by CLI flags for command-line use.

Conversion Defaults¶

ConversionConfig controls conversion and validation:

Field	Default	Meaning
`output_format`	`wav`	Converted format, `wav` or `flac`.
`sample_rate`	`16000`	Target sample rate.
`channels`	`1`	Target channel count.
`num_workers`	`4`	Parallel conversion workers.
`min_duration_sec`	`0.5`	Minimum valid output duration.
`overwrite`	`False`	Force regeneration of existing outputs.
`normalize_loudness`	`False`	Apply FFmpeg `loudnorm`.

CLI:

audio-prep convert \
    --input-dir data/raw_mp3 \
    --output-dir data/flac16k \
    --format flac \
    --sample-rate 16000 \
    --channels 1 \
    --workers 8

Python:

from audio_prep import ConversionConfig

config = ConversionConfig(
    output_format="flac",
    sample_rate=16_000,
    channels=1,
    num_workers=8,
)

Invalid values raise ValueError.

Chunking Defaults¶

ChunkConfig controls VAD chunking:

Field	Default	Meaning
`min_duration_sec`	`5.0`	Drop chunks shorter than this duration.
`max_duration_sec`	`20.0`	Split longer speech into windows at this duration.
`output_format`	`wav`	Chunk format, `wav` or `flac`.
`sample_rate`	`None`	Preserve native rate unless a target is supplied.
`num_workers`	`1`	Parallel chunking workers for Python API use.
`overwrite`	`False`	Force regeneration of valid existing chunks.
`allow_energy_fallback`	`False`	Use the energy detector if Silero cannot load.

CLI:

audio-prep chunk \
    --input-dir data/raw_mp3 \
    --output-dir data/chunks \
    --format flac \
    --sample-rate 16000 \
    --min-duration-sec 5 \
    --max-duration-sec 20 \
    --workers 4

Python:

from audio_prep import ChunkConfig

config = ChunkConfig(
    min_duration_sec=5,
    max_duration_sec=20,
    output_format="flac",
    sample_rate=16_000,
    num_workers=4,
)

The CLI sets --sample-rate 16000 for chunking by default. The Python API default is None, which preserves the source sample rate.