dvtag 0.7.7

A command-line tool designed to tag your doujin voice library.

DVTAG

Command-line tagger for doujin voice libraries. Point it at a folder tree that contains DLSite work numbers (RJ/BJ/VJ) in folder names and it will fetch metadata, write tags, and optionally transcode wav files.

What it does

• Recursively finds directories whose names include RJxxxxxx, BJxxxxxx, or VJxxxxxx (6 or 8 digits) and treats each as a release.
• Scrapes metadata from DLSite and refines title/cover details via Chobit when possible.
• Writes tags for FLAC, M4A, and MP3 (album, date, circle, seiyu, genres, cover, track/disc numbers).
• Tries to strip numbering prefixes from track titles based on filename patterns.
• Optional: transcode .wav to .flac or .mp3 using ffmpeg, deleting the source wavs after success.

Requirements

• Python 3.9+
• ffmpeg on PATH (only needed when using transcoding flags)
• Network access to dlsite.com and chobit.cc

Installation

pip install dvtag            # or: pipx install dvtag
pip install --upgrade dvtag  # upgrade (pipx upgrade dvtag)

Quick start

1. Ensure each release folder name contains a work number (e.g., RJ123456, rj123456 bonus, xxx_RJ01123456).
2. Run:

dvtag /path/to/your/library        # tag in place
dvtag -w2f /path/to/your/library   # tag + transcode wav -> flac
dvtag -w2m /path/to/your/library   # tag + transcode wav -> mp3 (320k)

CLI reference

usage: dvtag [-h] [-v] [-w2f] [-w2m] dirpath

Doujin Voice Tagging Tool (tagging in place)

positional arguments:
  dirpath        a required directory path

options:
  -h, --help     show this help message and exit
  -v, --version  show program's version number and exit
  -w2f           transcode all wav files to flac [LOSELESS]
  -w2m           transcode all wav files to mp3

How it works (and current quirks)

• Discovery: walks the given directory; once a folder with a work number is found, tagging happens inside it instead of recursing deeper.
• Metadata: pulls work name, circle, seiyu, genres, sale date, cover from DLSite; Chobit is queried for a potentially shorter title and a square thumbnail if available.
• Cover choice: currently prefers the Chobit square thumbnail, but some releases look better with the standard DLSite cover (usually landscape). A user-selectable preference is a known need.
• Genre order: genres are kept in the order returned by DLSite. The tag comparison logic may still rewrite files when the same genres appear in a different order; smarter equality is on the roadmap.
• Track titles: attempts to remove numeric prefixes from filenames when they follow common track-number patterns; otherwise uses the raw stem.
• Logging: minimal console logging; improving verbosity/structure is planned.
• Idempotency: tags are only written when they differ from existing tags, but the gaps above can still cause redundant writes.

Current limitations and ideas

• Caching: no cache; repeated runs on large libraries hit DLSite/Chobit every time, slowing execution and adding load. Add a local metadata cache with expiry.
• Genre/tag comparison: treat genre lists as sets (or configurable) to avoid unnecessary rewrites while preserving source order.
• Logging and observability: structured logs, progress, and clearer error handling.

License

MIT