Usage

silk has 5 subcommands:

silk <command> [options]

decode — SILK → WAV

silk decode <input.silk> [-o <output.wav>] [--to wav|pcm]

Default output is <input>.wav next to the input. --to pcm writes raw 16-bit little-endian samples (no WAV header).

encode — WAV → SILK

silk encode <input.wav> [-o <output.silk>] [--sample-rate 8000|12000|16000|24000]

WeChat SILK is typically 24kHz. The encoder uses the C SDK so output is bit-exact for re-sending via WeChat.

detect — magic-byte format detection

silk detect <file>

Output: file: format=<format> (extension: <ext>). Recognizes:

• #!SILK_V3 prefix → SILK (WeChat)
• RIFF...WAVE → WAV
• ID3 or 0xFF 0xEx → MP3
• OggS (+ Opus in page 2) → OGG or Opus
• fLaC → FLAC
• ADTS sync → AAC

info — file metadata

silk info <file>

For SILK and WAV, prints sample rate, channel count, sample count, duration, and bitrate. For other formats, prints format name and file size only.

batch — process a directory

silk batch <input_dir> [-o <output_dir>] [--to wav] [--pattern "*.silk"]

Default output dir is <input_dir>/converted/. Default pattern is * (all files). Non-audio files are silently skipped; already-target-format files are skipped (e.g., running batch --to wav on a directory with .wav files is a no-op).

Examples

Decode one file:

$ silk decode voice.silk -o voice.wav
decoded voice.silk (24000 Hz, 358560 samples) → voice.wav

Inspect:

$ silk info voice.silk
file: voice.silk
format: SILK (WeChat variant, 24000 Hz, mono)
samples: 358560
duration: 14.94s
size: 24812 bytes
bitrate: ~13.3 kbps

Batch convert 13 WeChat voice files in one go:

$ silk batch ./voices/ -o ./mp3s/ --to wav --pattern "*.silk"
✓ decrypted_01.silk → /tmp/out/decrypted_01.wav
... (12 more)
batch done: 13 converted, 0 errors