How do you turn JMdict or Jitendex into a Yomitan dictionary zip?
YuzuLingo · Nov 1, 2026
You either download a ready-made Yomitan zip built from JMdict or Jitendex, or you convert EDRDG XML yourself with yomitan-import and package the JSON as a format-3 zip. Most learners start from a published release instead of compiling from scratch.
Prefer a maintained release when you can
Projects such as jmdict-yomitan publish daily JMdict, JMnedict, and KANJIDIC zips that already match the Yomitan schemas. Jitendex is a further English glossary built from JMdict and Tatoeba with richer structured content. Pick the language edition you need, download the zip, and import it. Do not unzip and re-zip unless you are editing the banks.
Convert with yomitan-import
When you need a custom title, language flag, or a fresh EDRDG dump, use yomitan-import. Download a binary, point it at the JMdict XML source, choose the target language options, and write a destination zip. On Windows, keep non-ASCII characters out of EPWING source paths if you convert EPWING titles. The tool writes index.json and the term banks for you.
What a valid glossary zip looks like
The archive root must contain index.json with format 3, plus term_bank_*.json files. Optional tag banks and kanji banks can sit beside them. JSON must live at the root of the zip, not in a nested folder. Stored or Deflate compression is fine; Zip64 and loose .gz dumps are not a Yomitan dictionary.
JMdict versus Jitendex
Plain JMdict English zips cover the EDRDG glossary. Jitendex builds on that data with cleaner structured senses and examples. Many immersion setups use Jitendex as the primary English glossary and keep JMnedict or KANJIDIC as separate zips when names or kanji lookup matter. Other JMdict target languages still come from the multilingual JMdict releases.
After you have the zip
Import it in any tool that accepts Yomitan format 3 dictionaries. Pair it with separate pitch and frequency zips when you want accent graphs and corpus ranks. Rebuild or re-download when your release channel publishes a newer dump so deinflection tags and senses stay current.
