BlumeOps Docs

Rip a Disc

8 min read

Rip a Disc

Archive an audio CD, DVD or Blu-ray from the Pioneer BDR-S13U USB drive on indri into the navidrome or jellyfin library. Four mise tasks, run on indri (they need the drive), in two halves with a labeling step between them:

DiscExtractLabelFile
Audio CDmise run rip-cdedit metadata.jsonmise run rip-cd-finish <dir>
DVD / Blu-raymise run rip-videoedit metadata.jsonmise run rip-video-finish <dir>

The split is deliberate. Extraction is mechanical; naming is where the judgement lives — an obscure local pressing no database knows, a MusicBrainz candidate that is the wrong country’s release, CD-TEXT that dropped its apostrophes, or a TV disc where episode order has to be reconstructed from durations and chapter counts. That step is a JSON file anyone can edit, including an agent in a Claude Code session on indri, which is the intended way to run this: rip-cd, let the agent research and fill the draft, review, rip-cd-finish.

Audio CD

mise run rip-cd                    # → ~/rips/cd/<slug>-<discid8>/

rip-cd reads the table of contents and CD-TEXT with cd-info, computes the MusicBrainz and freedb disc IDs, asks MusicBrainz for releases matching the disc, then extracts with cd-paranoia (secure mode, full retries) and encodes to FLAC with ffmpeg. The drive ejects when extraction finishes. It writes:

  • 01.flac … NN.flac — untagged tracks
  • paranoia.log — cd-paranoia’s per-track summary; a blank bar is a clean read
  • rip.json — disc IDs, TOC, CD-TEXT, MusicBrainz candidates, and a submit_url for adding an unknown disc ID to MusicBrainz
  • metadata.json — the draft: an exact MusicBrainz match fills it in, otherwise CD-TEXT does, otherwise it is blank

Review metadata.json. artist, album and every track title are required; artist_sort names the library folder (Bowie, David); musicbrainz_release_id fetches cover art from the Cover Art Archive when cover (path or URL) is unset. Things the draft flags for you:

  • Several exact matches (_source says so, _candidates lists them) — the same album pressed in several countries or packagings. The first one filled the draft; swap musicbrainz_release_id if another is the disc in your hand, since that decides which cover art is fetched.
  • A disc of a set — disc_number/disc_total come from MusicBrainz when the release has more than one medium; set them by hand otherwise. The finish task files every disc of a set into the same album folder as <disc>-NN - Title.flac, so the discs can be ripped in any order.
  • A hidden track — audio of 2 s or more before track 1 is ripped as 00.flac and the draft carries a track 0 for it; title it, or delete both. Shorter pregap audio (common, and silent) is dropped without comment beyond a log line.

Then:

mise run rip-cd-finish ~/rips/cd/<dir> --dry-run   # show the plan
mise run rip-cd-finish ~/rips/cd/<dir> --clean     # tag, move, remove staging

Tracks land in /Volumes/music/<artist_sort>/<album>/NN - Title.flac with cover.jpg and <album>.rip.log (<album>.disc<N>.rip.log for a set) beside them, and navidrome finds them on its hourly scan (its NFS mount is the same sifaka share). Older albums in the library sit directly under the artist folder — that was XLD’s layout — and Navidrome does not care about the difference.

Re-running rip-cd on a disc whose staging dir exists resumes: tracks with a FLAC or a complete WAV are not re-extracted, a partial WAV from an interrupted run is discarded, only the missing span is ripped, and an edited metadata.json is never overwritten (a regenerated draft goes to metadata.draft.json beside it). --skip-rip refreshes rip.json without running cd-paranoia; it still needs the disc in the drive, since the disc IDs come from the TOC.

Audiobooks

An audiobook CD is ripped exactly like a music disc, but filed differently: set "kind": "audiobook" in metadata.json and the finish task writes Audiobookshelf’s layout, served by audiobookshelf, under the same share, one file per disc:

/Volumes/music/Audiobooks/<author>/[<series>/]<title>/Disc NN.opus

The metadata keys are artist (author), album (title), optional series, series_part and narrator, plus disc_number/disc_total. Track titles are not needed: audiobook CDs cut a track every few minutes as seek points for car players, not chapters, so the finish task concatenates the disc’s tracks losslessly and encodes once. Each disc then shows up as one chapter in Audiobookshelf; real chapters, if wanted, are added there by hand. Opus at 48 kb/s is the default because it is speech (a 17-hour book is about 400 MB rather than 5 GB); --lossless keeps FLAC.

MusicBrainz rarely knows audiobook discs, so expect to fill the draft by hand. Rip the discs in any order; disc_number names the file. The Audiobooks/ folder carries a .ndignore marker, so navidrome skips it and Audiobookshelf’s library is pointed at that folder alone.

DVD / Blu-ray

mise run rip-video                 # → ~/rips/video/<label>-<timestamp>/

rip-video inventories the disc with makemkvcon info, rips every title at least --min-length seconds long (default 600; --titles 0,3 picks explicitly), and ejects. rip.json lists each title’s duration, chapter count, size and source playlist; metadata.json is the draft, with the longest title pre-marked as the feature.

The inventory is makemkvcon’s own numbering (it hides titles under two minutes itself), and the staged files carry those ids — DD1_t06.mkv is title 6 — even though makemkvcon renumbers internally when it rips with a minimum length; the task renames its outputs back. A DVD’s play-all featurette usually appears once as a long title and again as its segments, each a few minutes; keep the long one.

Set kind:

  • movie — title, year; each file’s target is "feature", "extra:<Name>", or null to drop it.
  • tv — show, season; each file’s target is "E3", "E3:Episode Title", "E3-4" for a double episode, or null. season_dir overrides the Season NN folder name for shows already filed with another convention. Bonus-disc material takes "extra:<Name>" (filed under the season’s extras/) or "extra:<kind>/<Name>" with a Jellyfin extras folder name — featurettes, deleted scenes, behind the scenes, interviews, trailers — and shows up under that season’s Extras in Jellyfin with no sidecar metadata. Rip a bonus disc with --min-length 60, since most of its titles are shorter than the default cutoff.
mise run rip-video-finish ~/rips/video/<dir> --dry-run
mise run rip-video-finish ~/rips/video/<dir> --clean

Files go to /Volumes/allisonflix/Movies/<Title (Year)>/<Title (Year)>.mkv (extras under extras/) or /Volumes/allisonflix/TV/<Show>/Season NN/<Show> - SNNENN - <Title>.mkv, which jellyfin parses without any sidecar metadata. Nothing in the library is ever overwritten; a clash aborts the whole move.

A disc that will not read

Counterfeit pressings (shrink-wrapped, InterActual files and all) show up as MEDIUM ERROR: L-EC UNCORRECTABLE ERROR in makemkvcon’s messages, then Failed to save title N. Cleaning rarely helps; the sectors are bad in the plastic. Two things to know:

  • The drive wedges. On these sectors the Pioneer retries inside its firmware for minutes, the reading process sits in an uninterruptible wait at zero CPU, and afterwards drutil status and diskutil eject hang too. Kill the reader, then press the drive’s eject button or replug its USB cable. Nothing software-side clears it.

  • Image it, then rip the image. brew install ddrescue, unmount the volume (diskutil unmount force /Volumes/<label>), and run

    ddrescue -b 2048 -s <bytes from diskutil info> -n /dev/rdiskN disc.iso disc.map

    It copies everything readable in one pass and skips the rest, so a disc makemkvcon gives up on comes out 98% intact. Skip the retry and scrape passes (-r, no -n): each unreadable sector costs the drive 10–30 s, so recovering a few more kilobytes takes hours. Then makemkvcon -r mkv iso:disc.iso <title> <dir> extracts titles from the image with no drive involved, and a hand-written metadata.json (same schema as the draft) lets rip-video-finish file them. Decode-check the result (ffmpeg -v error -i f.mkv -map 0:v:0 -f null -, ignoring the null muxer’s non monotonically increasing dts noise, which every DVD rip produces) and note where the glitches fall before filing.

What this replaced

Until 2026-09 the drive was driven by GUI apps: XLD auto-started secure rips of audio CDs into ~/tmp (gnudb lookups, AccurateRip), MusicBrainz Picard tagged them, MakeMKV’s GUI ripped video into ~/Movies, and FileBot renamed it for Jellyfin. None of it was scriptable — XLD’s command line only decodes files, not discs — and FileBot’s license lapsed in 2025-11. The apps are still installed and still work for a one-off; the tasks above are the supported path.

Trade-off worth knowing: cd-paranoia verifies by re-reading, not against AccurateRip. For a pressing that is in AccurateRip and matters, XLD’s GUI still gives the stronger guarantee.

Requirements on indri

ToolSourceUsed by
cd-paranoia, cd-infoHomebrew libcdio-paranoia (in the dotfiles Brewfile)rip-cd
ffmpegHomebrewrip-cd
makemkvconHomebrew cask makemkv (needs a current MakeMKV key)rip-video
drutilmacOSboth
/Volumes/music, /Volumes/allisonflixsifaka-mounterthe finish tasks

One process on the drive at a time. A makemkvcon info probe while cd-paranoia is reading (or the MakeMKV GUI opening on insert) contends for the drive and can leave the rip stuck in an uninterruptible read with zero CPU and a WAV that stops growing. Kill it and rerun; the rip resumes. Disc auto-launch is off on indri (defaults read com.apple.digihub shows action = 1 for music CDs and video DVDs); keep it that way.

The tasks guard their binaries with _require, so running one from gilbert or an agent pod fails at the door with an explanation.

Graph View

Backlinks