Skip to content

SUPPORTED_FORMATS.md

Latest commit

 

History

History
110 lines (73 loc) · 3.44 KB

SUPPORTED_FORMATS.md

File metadata and controls

110 lines (73 loc) · 3.44 KB

English | 中文

Supported Audio Formats

Decoder Routing

The tool uses smart auto-routing, preferring Symphonia with automatic FFmpeg fallback when needed.

Native Symphonia Support

  • Lossless: FLAC, ALAC (Apple Lossless), WAV, AIFF, PCM
  • Lossy: AAC, OGG Vorbis, MP1 (MPEG Layer I)
  • Containers: MP4/M4A (Symphonia-supported codecs only), MKV/WebM

Dedicated Decoders

  • Opus: Via songbird decoder (Discord audio library)
  • MP3: Stateful format, forced serial decoding

Auto Fallback to FFmpeg

When Symphonia cannot decode a format, the tool automatically falls back to FFmpeg.

Typical cases:

  • For extensions .ac3, .ec3/.eac3, .dts, .dsf, .dff → use FFmpeg directly
  • MP4/M4A containers with AC-3/E-AC-3 (incl. Atmos) or DTS → auto-switch to FFmpeg
  • Incompatible codecs inside containers (some MKV/MP4 variants) → auto fallback to FFmpeg

FFmpeg Installation

To use FFmpeg features, make sure both ffmpeg and ffprobe are installed:

Platform Install Command
macOS brew install ffmpeg
Windows winget install Gyan.FFmpeg (or Chocolatey)
Ubuntu/Debian sudo apt install ffmpeg
Fedora/RHEL sudo dnf install ffmpeg
Arch sudo pacman -S ffmpeg

Verify: both ffmpeg -version and ffprobe -version should print a version; the tool auto-detects them from PATH.

Multichannel & LFE Support

  • Multichannel analysis: supports 3–32 channels; per-channel DR is computed and listed.

  • Official aggregation: arithmetic mean of all non-silent channel DRs, rounded (foobar2000 style).

  • LFE detection:

    • Via Symphonia: auto-detects layout metadata (e.g., WAV WAVEFORMATEXTENSIBLE masks, some MP4/MKV).
    • Via FFmpeg: parses ffprobe JSON label sequences (e.g., FL+FR+FC+LFE+…) to locate LFE accurately.

  • LFE exclusion (optional): enable --exclude-lfe to drop LFE from the aggregate; per-channel DR lines remain.

DSD Processing

Options

Flag Description Default
--dsd-pcm-rate Target sample rate 352800 Hz
--dsd-gain-db Linear gain (0 to disable) +6.0 dB
--dsd-filter Low-pass filter teac

Filter Modes

teac (TEAC Narrow):

  • DSD64 → 39 kHz
  • DSD128 → 78 kHz
  • DSD256 → 156 kHz
  • DSD512 → 312 kHz
  • DSD1024 → 624 kHz
  • Capped at 0.45×Fs (target rate)

studio:

  • Fixed 20 kHz (audible-band only)

off:

  • No extra low-pass (diagnostic; ultrasonics enter RMS and may reduce DR; clipping risk with +6 dB)

Output Format

  • Unified F32LE output for consistency and easy processing
  • Reports show "native 1-bit rate & tier → processed rate"; bit depth printed as "1 (DSD 1-bit, processed as f32)"

Format Summary

12+ mainstream formats, covering 90%+ user needs:

Category Formats Decoder
Lossless FLAC, ALAC, WAV, AIFF, PCM Symphonia
Lossy AAC, OGG Vorbis, MP1 Symphonia
Proprietary MP3 Symphonia (Serial)
Proprietary Opus songbird (Dedicated)
Video Codec AC-3, E-AC-3, DTS, DSD FFmpeg (Auto)
Containers MP4/M4A, MKV, WebM Symphonia / FFmpeg (Smart)

Parallelism Notes

  • MP3 uses serial decoding (stateful format); other formats support parallel acceleration.
  • Multichannel uses zero-copy strided optimization with 8–16× performance gain for 3+ channels.