pydsscodec

Python bindings for the Rust dss-codec decoder.

pydsscodec decodes Olympus DSS and DS2 audio, including Grundig/PH9607 DSS, directly in Python through a native Rust extension. It mirrors the core dss-codec model: top-level decode and decrypt functions plus streaming decoders.

Features

• Decode DSS SP, Grundig DSS SP, DS2 SP, DS2 QP, and DS2 QP7 files from paths or bytes
• Normalize encrypted DS2 input back to plain container bytes
• Inspect format, native sample rate, and encryption metadata without decoding
• Stream decode in chunks without forcing a whole-file convenience path
• Load native Rust code in-process instead of shelling out to a CLI
• In our benchmark, achieved roughly 150x faster DSS decoding and 50x faster streamed DS2 decoding than the reference Python decoders

Installation

pip install pydsscodec

If a prebuilt wheel is not available for your platform, installation will build the Rust extension locally.

Prebuilt wheels are currently published for:

• Linux x86_64
• macOS arm64
• Windows x86_64

Other platforms install from source and require Rust 1.85 or newer.

Usage

Decode a file:

from pydsscodec import decode_file

audio = decode_file("recording.ds2")
print(audio.format, audio.sample_rate, audio.native_rate, audio.duration_seconds)
print(audio.sample_count)

Decode in-memory bytes:

from pydsscodec import decode_bytes

audio = decode_bytes(data)

Decrypt encrypted DS2 input to plain container bytes:

from pydsscodec import decrypt_file

plain_ds2 = decrypt_file("encrypted.ds2", password="secret")

Stream decode in chunks:

from pydsscodec import DecryptingDecoderStreamer

decoder = DecryptingDecoderStreamer(password="secret")

with open("recording.ds2", "rb") as handle:
    while chunk := handle.read(4096):
        chunk_samples = decoder.push(chunk)
        # process chunk_samples here

final_samples = decoder.finish()

print(decoder.format(), decoder.native_rate(), len(final_samples))

Detect the container format without decoding:

from pydsscodec import detect_format

fmt = detect_format(data)
print(fmt)  # e.g. "dss_sp", "grundig_sp", "ds2_sp", "ds2_qp", or "ds2_qp7"

Inspect format and encryption metadata without decoding:

from pydsscodec import inspect_file

info = inspect_file("recording.ds2")
print(info.format, info.native_rate, info.encryption, info.encryption_mode)

API

Top-level functions:

• decode_file(path, password=None) -> DecodedAudio
• decode_bytes(data, password=None) -> DecodedAudio
• decrypt_file(path, password=None) -> bytes
• decrypt_bytes(data, password=None) -> bytes
• detect_format(data) -> str | None
• inspect_file(path) -> FileInfo
• inspect_bytes(data) -> FileInfo

Streamer classes:

• StreamingDecoder()
• DecryptStreamer(password=None)
• DecryptingDecoderStreamer(password=None)

Streamer methods mirror the Rust API:

• push(data)
• finish()
• format()
• native_rate()

Bulk decode returns a DecodedAudio object with:

• samples: decoded mono samples as list[float]
• sample_rate: sample rate of the returned samples
• native_rate: original sample rate of the source codec
• format: one of "dss_sp", "grundig_sp", "ds2_sp", "ds2_qp", or "ds2_qp7"
• sample_count
• duration_seconds

Inspection returns a FileInfo object with:

• format: detected audio format
• native_rate: native sample rate in Hz
• encryption: "none", "ds2_aes128", "ds2_aes256", or "unknown"
• encryption_mode: the numeric DS2 encryption mode, or None for plain input

Notes

• The streaming classes are the preferred scalable interface for large inputs.
• Bulk decode returns samples as a Python list.