The Grey Pilgrim speaks many tongues. A C++17 CUE sheet splitter with tagging and MusicBrainz Disc ID computation — the proper successor to the gandalf bash function.

Reads a CUE sheet (or extracts an embedded one from FLAC), splits the source audio into individual tracks using ffmpeg, tags them via TagLib, and optionally computes a MusicBrainz Disc ID for easy release lookup.

• ffmpeg / ffprobe — audio splitting and duration probing
• TagLib (≥ 1.12) — metadata writing (FLAC Xiph comments, generic tags)
• OpenSSL (≥ 1.1) — SHA-1 for MusicBrainz Disc ID
• metaflac (optional) — fallback for embedded CUE extraction
• CMake ≥ 3.16, a C++17 compiler

sudo dnf install cmake gcc-c++ taglib-devel openssl-devel ffmpeg metaflac

sudo apt install cmake g++ libtag1-dev libssl-dev ffmpeg flac

mkdir build && cd build
cmake ..
make -j$(nproc)

The binary lands at build/mithrandir.

./mithrandir album.cue

Outputs FLAC files (compression 8) alongside the source, named 01 - Track Title.flac, etc. Tags are written automatically.

./mithrandir album.flac

Extracts the embedded CUE sheet, then splits and tags as above.

./mithrandir -o ~/Music/split -p "{number}. {artist} - {title}" album.cue

Available tokens: {number}, {title}, {artist}, {album}, {date}, {genre}.

./mithrandir -f wav album.cue

./mithrandir --dry-run -v album.cue

Shows what would be done (timestamps, filenames) without writing anything.

./mithrandir --pregap prepend album.cue   # include pregap in each track
./mithrandir --pregap append album.cue    # append pregap to previous track
./mithrandir --pregap discard album.cue   # default — start at INDEX 01

When splitting from a single-FILE CUE (typical for CD rips), mithrandir computes a MusicBrainz Disc ID from the TOC and prints a lookup URL:

MusicBrainz Disc ID: dX5mFkM7h3lKsNBIbcfN4gJfEOo-
  Lookup: https://musicbrainz.org/cdtoc/dX5mFkM7h3lKsNBIbcfN4gJfEOo-
  TOC:    https://musicbrainz.org/cdtoc/attach?toc=1+12+...&tracks=12

The TOC URL works even if the Disc ID isn't in the MusicBrainz database yet — it searches by track offsets, which often matches the correct release.

main.cpp          CLI argument parsing, orchestration
types.h / .cpp    Shared data model (CueSheet, CueTrack, SplitOptions)
cue_parser        CUE parsing (file + string), embedded CUE extraction
splitter          ffmpeg-based track extraction with pregap handling
tagger            TagLib-based metadata writer (Xiph comments for FLAC)
discid            MusicBrainz Disc ID computation (SHA-1 + custom base64)

• TagLib over metaflac: TagLib gives native C++ access to Xiph comments (ALBUMARTIST, DISCNUMBER, TOTALTRACKS, ISRC, BARCODE) without shelling out. It also handles WAV/WV/OGG if you ever need those.

• ffmpeg for splitting: Sample-accurate seeking for audio, handles FLAC, WAV, WavPack (including 32-bit float), APE, and anything else ffmpeg supports. -ss after -i ensures precise timestamps.

• Custom CUE parser: CUE is simple enough that a custom parser gives more control over edge cases (multi-FILE, REM extensions, encoding quirks) than libcue, with zero extra dependencies.

• Compression level 8: Matches your gimli convention. Maximum FLAC compression for archival; negligible speed difference on modern hardware.

• Character encoding: Currently assumes UTF-8 (or Latin-1 subset). CUE files from Japanese CDs (Shift-JIS) need manual conversion via iconv first. A future version could auto-detect via ICU or uchardet.

• Cover art: Not handled. Use MusicBrainz Picard or metaflac --import-picture-from post-split.

• ReplayGain: Not computed. Run metaflac --add-replay-gain *.flac on the output directory after splitting.

• Parallel splitting: Tracks are split sequentially. For large albums on NVMe, parallel extraction via std::async would reduce wall time.

MIT