A desktop companion for digitising vinyl records via Audacity — silence detection, Discogs metadata, waveform editing, and one-click tagged export.

VRipr sits alongside Audacity while you rip a vinyl record. It:

1. Connects to Audacity via the scripting pipe and exports the current project to a clean analysis WAV — capturing any edits you've already made (needle-drop removal, fades, etc.)
2. Fetches the album's track listing and durations from Discogs
3. Detects track boundaries using one of three algorithms — RMS energy, Spectral flatness, or an adaptive HMM — with multi-pass retries and Discogs-guided anchoring when durations are available
4. Shows an interactive waveform with draggable track markers and a right-click context menu for auditioning and pinning boundaries in Audacity
5. Sets labels in Audacity for a final visual review, then exports each track as a tagged FLAC/MP3/WAV/OGG with full metadata including multi-value GENRE and ARTIST tags, a DISCOGS_RELEASEID custom tag, and folder.jpg cover art

The waveform panel shows the full recording with coloured track regions and draggable boundaries. The track table carries Discogs metadata. Cover art is displayed on the right.

Progress indicator, per-track log output, and export paths are shown in real time.

Search results include label, format, year, and a 🌐 button to open the release page directly in your browser. Hover over any row for full release details.

After clicking Set Labels, Audacity shows a label track with the detected track boundaries. Drag them in Audacity if needed, then switch back to VRipr and click Export All.

1. Open Audacity → Edit → Preferences → Modules
2. Set mod-script-pipe to Enabled
3. Restart Audacity

VRipr communicates with Audacity through named pipes. If the connection fails, check the Diagnostics button — it reports the expected pipe paths.

GitHub Actions builds a native binary for every push to main and on every v* tag:

Download from the Releases page. All binaries are self-contained — no runtime dependencies, no system libraries required.

git clone https://github.com/yourname/vripr.git
cd vripr
cargo build --release
./target/release/vripr

The release binary has LTO enabled and is self-contained. Copy it anywhere on your PATH.

# Install binary
sudo install -Dm755 target/release/vripr /usr/local/bin/vripr

# Desktop entry
cat > ~/.local/share/applications/vripr.desktop <<EOF
[Desktop Entry]
Name=VRipr
Comment=Master Vinyl Rippage
Exec=vripr
Icon=/path/to/assets/vripr.webp
Type=Application
Categories=Audio;
EOF

Config is stored at ~/.config/vripr/vripr.toml (created automatically on first save).

VRipr normalises genre tags on export using a mapping table (genre.dat). The built-in table ships with ~640 entries drawn from the sanitizegenre project and covers:

• Abbreviations — HH → Hip-Hop;Hip Hop, DT → Dub Techno, E → Electronic …
• Typo corrections — Trchno → Techno, Popo → Pop, Shoegzae → Shoegaze …
• Case normalisation — TECHNO → Techno, HOUSE → House …
• Multi-genre expansion — Folk Pop → Folk Pop;Folk;Pop, Prog Rock → Prog Rock;Prog-Rock;Progressive Rock;Progressive …
• Discogs styles — when a release is fetched, both the genres[] and styles[] arrays are combined into a semicolon-delimited string (e.g. Electronic;Deep House;House) so the full genre hierarchy is preserved through to the exported tags

On export each semicolon-delimited component is looked up, expanded, deduplicated, and written as one GENRE tag per value. Players that support multi-value tags (foobar2000, beets, Picard, Kodi, Plex…) will display and filter on all genres. Single-value players receive the first genre.

You can supply your own genre.dat to match your personal library conventions, language preferences, or custom taxonomy. The format is one mapping per line:

# comment lines start with #
SourceString|Target1;Target2;Target3

• Source is the raw string that appears in the genre field (exact match first, case-insensitive fallback).
• Target is a semicolon-separated list of canonical output values.
• Unknown genres pass through unchanged — you only need to list the exceptions.

Example entries:

# French genre names
Électronique|Electronic
Jazz Fusion|Jazz Fusion;Jazz
Hip Hop|Hip-Hop;Hip Hop;Rap

# Studio-internal codes
EXP|Experimental;Electronic
CJAZZ|Contemporary Jazz;Jazz

To activate a custom file: open Settings → Genre Map File, click …, and select your .dat file. The map is loaded immediately on Save and persisted to vripr.toml. Click ✕ in the field to revert to the built-in mappings.

A release may credit multiple artists on a single track, or an artist may be known by more than one name. VRipr supports this by treating the Artist field as a semicolon-delimited list:

Daniel Mana;Mana

On export, each entry is written as a separate ARTIST tag — players that support multi-value tags (foobar2000, beets, Picard, Kodi, Plex…) will index and display all contributors. The Album Artist field is treated identically.

Enter multiple artists in the edit panel using ; as the delimiter. No sanitisation map is applied — the values are written exactly as entered after splitting and trimming whitespace.

Rip your vinyl into Audacity as normal. Remove the needle drop and any leading/trailing silence if you wish — VRipr exports the project at connect time, so your edits are included.

Click Connect in the toolbar. VRipr opens the Audacity scripting pipe, exports the full project to a temporary WAV, and builds the waveform display. This takes 30–90 seconds for a full LP side.

Fill in Artist and Album in the Apply-to-all strip and click 📀 Fetch Release. Choose the correct release from the Discogs picker — use the 🌐 button to verify on the Discogs website if unsure.

VRipr then runs track detection (see Detection algorithm below). The log panel shows what happened at each step.

If the release has no track durations (titles only), VRipr still populates the track table with placeholder entries — all timings are set to zero and can be adjusted manually in the edit panel.

The track table's Time column shows start–end (duration) at a glance.

The waveform panel shows coloured regions for each detected track. The amplitude colour gradient — teal for quiet passages, through green and yellow to red at loud peaks — gives an immediate sense of the dynamic shape of each track. If any boundary is wrong:

• Drag the boundary line left or right to adjust it. Adjacent track boundaries move together.
• Right-click anywhere in the waveform to open the context menu:
  • ▶ Play Track N — sends a SelectTime: + Play: command to Audacity to audition that region
  • ↦ Pin start here / ↤ Pin end here — moves the boundary to the click position
  • 📌 Pin/Unpin — locks a track so it is preserved on re-scan
  • ⏹ Stop playback — stops Audacity playback
• Drag empty space to draw a selection, then click ➕ Add Track to insert a new track at exactly that position.
• Click 🔄 Re-scan to run detection again while keeping pinned tracks.

Double-click any row in the track table to open the edit panel. Three time fields are available:

All three fields accept MM:SS.ss notation and display it consistently after editing.

One side per Audacity session is strongly recommended. The silence detector works on whatever audio is loaded — if you record multiple sides as a single continuous file, it must find twice (or four times) as many tracks with no knowledge of where the physical side-break is. Detection reliability drops significantly above ~6 tracks. For double albums, triple albums, and boxed sets, processing one side per session is not just a suggestion — it is effectively required for consistent results.

The Side: selector in the toolbar handles the metadata side of this workflow. It appears automatically when the loaded Discogs release has more than one vinyl side. For double albums (4 sides) and above it groups sides into discs — Disc 1 / Side A, Disc 1 / Side B, Disc 2 / Side C — so you always know which disc a side belongs to.

Workflow for a double album (4 sides, 2 discs):

1. Record Disc 1 Side A into Audacity. Connect → detect → select Side A (Disc 1) → export. Done for Side A.
2. New Audacity session. Record Disc 1 Side B. Connect → detect → select Side B (Disc 1) → export.
3. Repeat for Disc 2 Sides C and D.

Ripping out of order (e.g., Side B before Side A):

1. Load your Side B recording and connect.
2. Fetch the release. Select Side B from the combo.
3. VRipr filters the Discogs tracklist to Side B only and calibrates the retry loop against that side's track count.
4. If you change the side selection after detection has already run, VRipr immediately re-assigns Discogs metadata to the existing detected boundaries without re-running audio analysis. Click 🔄 Re-scan to redo detection calibrated to the new side's expected count.

Track numbers are assigned from the Discogs vinyl position (B1, B2, …) regardless of the order you rip, so exported files sort correctly in your library.

What the Side selector does not do: it does not make multi-side audio easier to analyse. If you've recorded Sides A and B as one 90-minute file, selecting "Side A" only filters the metadata — the detector still has to find all the tracks in 90 minutes of audio. The best approach is to rip all sides in sequence as a single continuous recording, then use the All option in the Side selector — VRipr will detect tracks across the full recording and match them to the complete tracklist in order.

Click 🏷 Set Labels. VRipr clears any existing label track in Audacity, writes new labels from the track table, then reads them back to confirm they took. Switch to Audacity for a final visual check — drag labels if needed.

If any track boundary needs adjusting, Audacity's label editor gives you a visual waveform to work against:

1. Click 🏷 Set Labels to push the current boundaries into Audacity.
2. Switch to Audacity. Drag the label edges left or right until each boundary sits exactly at the silence gap.
3. Switch back to VRipr and click ⬇ Get Labels. VRipr reads the adjusted positions back and updates the track table in place — all metadata (title, artist, album, etc.) is preserved, only the start/end times change.
4. Repeat as needed. Set Labels → tweak in Audacity → Get Labels is a fast loop with no re-detection required.

This is particularly useful for albums with ambiguous silence gaps, mid-track fades, or live recordings where applause bleeds into the next track.

Click 💾 Export All. VRipr:

• Validates that label count and titles match (non-blocking warning if not)
• Selects each time region in Audacity and exports it via Export2:
• Writes full ID3/Vorbis/FLAC tags including DISCOGS_RELEASEID, multi-value GENRE and ARTIST
• Deposits folder.jpg in the album directory after the first track completes

Output structure follows your Path Template. The default template produces:

{Export Directory}/
  {Album Artist}/
    {Album}/
      folder.jpg
      01 - Track Title.flac
      02 - Track Title.flac
      …

The Path Template setting (under Export & Detection in Settings) controls where each exported file is placed relative to the Export Directory.

Wrap tokens in [...] to make them conditional: if the token resolves to an empty string, the entire [...] group (including brackets) is removed. This avoids orphaned [] in directory names when metadata is missing.

Default:

{album_artist}/{album}/{tracknum} - {title}

With ISO country code and catalogue number:

{album_artist}/{album} [{country_iso}][{catalog}]/{tracknum} - {title}

→ Miles Davis/Kind of Blue [US][CL 1355]/01 - So What.flac → (if country is empty) Miles Davis/Kind of Blue [CL 1355]/01 - So What.flac → UK release: Miles Davis/Kind of Blue [GB][CL 1355]/01 - So What.flac

The Album Name Format setting (also under Export & Detection) lets you customise the Album metadata tag written to every exported file using the same {token} syntax as the Path Template. Leave it blank and the album name from Discogs (or whatever you've typed in) is used verbatim. Set a format and it is expanded per-track at export time.

Useful for collectors who want release context baked directly into the tag rather than only into the folder structure:

{album} [{country_iso}][{catalog}]

→ Kind of Blue [GB][CBS 62066]

{album} ({year})

→ Kind of Blue (1959)

Bracket collapsing applies here too — [{country_iso}] disappears cleanly if the country is unknown.

VRipr ships three track-boundary detectors plus a duration-guided pass. All share the same multi-pass retry loop.

When every expected track has a duration in the Discogs data, VRipr runs a guided detection first: it uses the Discogs durations as timing anchors, scanning only the narrow windows around where each track boundary is expected to be. This is faster and more reliable than a blind full-file scan when the timing data is good. If the guided pass produces the exact expected track count it is used directly; otherwise VRipr falls back to the blind retry loop below.

The classic energy-based scanner:

1. Decodes the analysis WAV to per-window RMS values (50 ms windows)
2. Applies hysteresis to avoid rapid toggling on borderline signals
3. Bridges small gaps (vinyl crackle, pops) shorter than the gap-fill threshold
4. Requires a minimum silence duration for a gap to count as a track boundary
5. Discards regions shorter than the minimum track duration

Works well for clean pressings where the inter-track groove is meaningfully quieter than the music.

A combined energy + spectral-flatness scanner, better suited to noisy pressings where the inter-track groove is loud but acoustically different from music.

The key insight: surface noise is spectrally flat (energy spread roughly evenly across all frequencies, like white noise) while music is spectrally peaked (energy concentrated in harmonically related bands). Spectral flatness — the ratio of the geometric mean to the arithmetic mean of the power spectrum — measures this distinction on a 0–1 scale: 0 = perfectly tonal, 1 = white noise. Inter-track groove noise typically scores 0.80–0.95; music 0.10–0.50.

For each 50 ms window the detector computes:

1. RMS — as in the energy-only scanner
2. Spectral flatness via FFT (Hann-windowed, zero-padded to the next power of two, positive-frequency half-spectrum). A ±2-window (≈250 ms) rolling average smooths the flatness signal before thresholding.

A frame is classified as between tracks if its RMS falls below the energy threshold (ordinary silence), or its flatness exceeds the flatness threshold while energy is still present (energetic surface noise).

When to use it: if your pressing is noisy, the RMS detector reports more tracks than expected, and lowering the threshold starts eating into quiet musical passages. The flatness threshold defaults to 0.85.

A two-state Hidden Markov Model over (RMS dB, spectral flatness) features. Unlike the threshold-based detectors, the HMM learns emission parameters from the recording itself:

• The bottom 15 % of frames by RMS are treated as silence examples → Gaussian model for the SILENCE state
• The top 40 % by RMS are treated as music examples → Gaussian model for the SOUND state
• Transition probabilities are derived from the configured minimum silence / minimum track duration

Viterbi decoding finds the globally most-likely sequence of SILENCE/SOUND states. Because the HMM assigns a cost to switching states, momentary level dips mid-track — a quiet passage, a sudden dynamic contrast — no longer produce spurious track splits. The same post-processing (gap-fill, minimum duration, de-overlap) runs on the extracted sound regions.

When to use it: when RMS and Spectral both split tracks incorrectly at quiet passages and manual threshold tuning isn't converging. The HMM needs no threshold configuration — it adapts to each recording.

Implementation: pure Rust, no additional dependencies. Uses the same FFT-based feature extraction as the Spectral detector.

If the detected count doesn't match Discogs, all detectors retry up to four more times, progressively shortening the minimum silence and minimum track duration each pass.

After all retries:

• VRipr uses Export2: via the scripting pipe — this requires Audacity 3.x with mod-script-pipe enabled.
• The pipe is synchronous: VRipr waits for each command to complete before proceeding.
• Waveform playback (▶ Play Track) sends SelectTime: + Play: commands to Audacity — Audacity must be open and connected for this to work.
• Labels are written with AddLabel: + SetLabel: and verified by reading them back with GetInfo: Type=Labels.
• The analysis WAV is written to /tmp/vripr_analysis_{pid}.wav and is safe to delete after a session.

Bug reports and pull requests are welcome. Please open an issue first for significant changes.

See CONTRIBUTING.md for build instructions, code structure, and conventions.

Team Badger shirts and other goodies are available at shunte88

MIT — see LICENSE.