disc-forge

Automated CD burning for macOS via cdrdao. Two modes:

• Audio CD (default) — Red Book, plays in any CD player, 80-min capacity, CD-Text from ID3 tags
• MP3 data CD (--mp3-disc) — plays in MP3-capable stereos (most 2005+ car head units), 700 MB capacity, ISO9660+Joliet filesystem

Turns "burn the BSG soundtrack to a CD" or "make an MP3 disc of these mantras" into a single command.

When to Use

• User asks to burn an album, soundtrack, or playlist to an audio CD for a car stereo / home stereo
• Source music lives on a remote media host (reached over SSH) or in a local folder (this machine's host and album paths: ~/.claude/agent_docs/music.md, if present; see references/media-layout.md)
• A USB CD burner is plugged into the current Mac
• User wants either a Red Book audio CD or an MP3 data CD

Out of scope: DVD burning, CD ripping (use XLD — see ~/.claude/agent_docs/music.md), multi-session discs, DDP mastering, mixed-mode discs (audio + data on one disc).

Quick Start

Burn a full album from the media host with one command:

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source "media-host:/Volumes/Music Library/BSG.S1 - OST/" \
  --name BSG-S1

That runs the full pipeline: preflight → tar-over-SSH stage → duration check → parallel WAV conversion → TOC build with CD-Text → cdrdao write at 16x → cleanup → eject. Takes ~5–7 minutes for an 80-minute disc. cdrdao streams its own progress to stdout; a 10-second countdown precedes the actual burn so Ctrl-C still aborts safely.

Before the first burn, make sure a blank writable CD-R or CD-RW is in the drive — preflight.py will abort early if not.

The Pipeline

burn_audio_cd.py is the master orchestrator. It chains the stage scripts in sequence and halts on any failure, preserving staging for debugging. Each script is also independently usable.

Shared stages (both modes):

Audio CD stages (default):

MP3 data CD stages (--mp3-disc):

Scripts

All scripts support --help. All live under ~/.claude/skills/disc-forge/scripts/. Default staging root is ~/burn-staging/<name>/.

preflight.py         # preflight.py [--json]
stage_tracks.py      # stage_tracks.py --source <path|host:/path> --dest <name> [--format mp3]
check_duration.py    # check_duration.py <staging-dir> [--ceiling 80] [--json]
convert_to_wav.py    # convert_to_wav.py <staging-dir> [--output <wav-dir>] [--jobs N]
build_toc.py         # build_toc.py --wav-dir <wavs> --source-dir <mp3s> --output <toc> [--gaps]
build_data_toc.py    # build_data_toc.py --source-dir <dir> --volume-label <label> --iso-output <iso> --toc-output <toc>
burn_audio_cd.py     # burn_audio_cd.py --source <path|host:/path> --name <name> [--mp3-disc] [--speed 16] [--gaps] [--dry-run] [--keep-wavs] [--ceiling 80]

Key burn_audio_cd.py flags:

• --mp3-disc — burn an MP3 data CD (ISO9660+Joliet) instead of a Red Book audio CD. 700 MB capacity, plays in MP3-capable car stereos (most 2005+ head units).
• --data-ceiling N — MP3 data disc capacity ceiling in MB (default 700, only with --mp3-disc).
• --dry-run — runs everything through TOC generation, skips the cdrdao write. Use to validate the whole pipeline on a disposable run.
• --gaps — 2-sec spacers between tracks (pop albums). Default is gapless (right for OSTs, film scores, live albums). Audio mode only.
• --speed N — cdrdao write speed multiplier. Default 16. Drop to 10 for maximum reliability on cheap media.
• --keep-wavs — preserve the intermediate WAV staging dir (audio mode) or ISO (data mode) after the burn.
• --ceiling N — audio CD capacity in minutes for the fit check. Default 80. Set to 74 for older 74-min media.

Recipes

Burn a full soundtrack from the media host (the canonical path):

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source "media-host:/Volumes/Music Library/Blade Runner - OST/" \
  --name BladeRunner-OST

Burn from a pre-staged local directory (when tracks are already on the laptop):

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source ~/burn-staging/BSG-S1 \
  --name BSG-S1

The orchestrator detects that source == staging dir and skips the stage step.

Dry-run the pipeline (validate without burning a coaster):

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source ~/burn-staging/BSG-S1 \
  --name BSG-S1 \
  --dry-run

Runs every stage through TOC generation. Prints the cdrdao command you'd run next and the cdrdao show-toc command to validate the generated TOC.

MP3 data CD for a modern car stereo (2005+ head units, ~700 MB capacity):

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source ~/burn-staging/mantras \
  --name mantras \
  --mp3-disc

Volume label derives from --name (sanitized to uppercase A-Z/0-9/_). Stage MP3s with numeric prefixes (01 Foo.mp3, 02 Bar.mp3) to force track order — most stereos play alphabetically. For MP4/Opus/M4A sources, transcode to MP3 first (stage_tracks.py filters to .mp3 by default).

Pop album with 2-sec gaps (Led Zeppelin, Pink Floyd, etc.):

python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py \
  --source "media-host:/Volumes/Music Library/Dark Side Of The Moon/" \
  --name DSOTM \
  --gaps

Cherry-picked mixtape (tracks from multiple albums):

Stage tracks manually into a single folder first, then point the orchestrator at it:

mkdir -p ~/burn-staging/my-mixtape
scp "media-host:/Volumes/Music Library/Blade Runner - OST/01 Main Titles.mp3" ~/burn-staging/my-mixtape/
scp "media-host:/Volumes/Music Library/Inception OST [2010]/01 Half Remembered Dream.mp3" ~/burn-staging/my-mixtape/
# ... rename files with numeric prefixes to force desired order ...
python3 ~/.claude/skills/disc-forge/scripts/burn_audio_cd.py --source ~/burn-staging/my-mixtape --name my-mixtape

Track order is determined by sorted() on filenames — use numeric prefixes (01_, 02_, …) to force a specific sequence.

Critical Reminders

Full expansions (with why) live in references/gotchas.md.

• SMB is not reachable over Tailscale. Port 445 is refused on Tailscale addresses; only port 22 works. Use SSH transport. stage_tracks.py already does this.
• Apple's bundled rsync (openrsync 2.6.9) splits remote paths on spaces. stage_tracks.py uses tar-over-SSH instead. Never use rsync for remote paths with spaces.
• cdrdao 1.2.6 rejects MP3 input directly. The binary contains the literal string "AIFF and MP3 not supported by cdrdao". WAV conversion is mandatory — convert_to_wav.py handles it. (Burn.app is the alternative if you want to skip WAV conversion — see references/burn-app-fallback.md.)
• drutil "Unsupported" label is a red herring for cdrdao. It only affects Apple's DiscRecording framework (Finder burn / Music.app burn / Burn.app), not cdrdao. preflight.py surfaces the label as a warning, not an error.
• CD-Text must come from source ID3 tags, not WAVs. PCM WAV has no tagging container. build_toc.py reads tags from the original file, references the matching WAV in AUDIOFILE.
• macOS Sequoia killed Music.app and Finder's burn-to-disc. Don't fall back to them.

References

Dependencies

• cdrdao — brew install cdrdao
• ffmpeg + ffprobe — brew install ffmpeg
• SSH access to the media host (only when sourcing remotely)
• A USB CD burner that speaks MMC. The ASUS SDRW-08U9M-U works despite drutil labeling it Unsupported.