How do I import dictionaries (Yomitan / MDX)? Which kinds do I need for Japanese?
Fushi accepts these dictionary formats, all imported as-is with no conversion:
- Yomitan (Yomichan) dictionary zips — the common format of the browser lookup ecosystem; every dictionary collection prepared for Yomitan works in Fushi;
- MDict
.mdx— the format used by the huge library of GoldenDict / Eudic / DictTango dictionaries; ones with images or audio come with a same-named.mdd, select it too; - Lingvo
.dsl; - StarDict.
Yomitan dictionaries have the richest structured fields (pitch accent, frequency, inflections); MDX wins on sheer volume — many monolingual and bilingual dictionaries only exist as MDX.
What the recommended pack already gives you
After installing the recommended pack (about 9.5 GB) from onboarding you have:
- Japanese vocabulary dictionaries (Japanese–Japanese / Japanese–Chinese);
- a pitch-accent dictionary (the accent marks in the lookup popup);
- frequency dictionaries (the frequency badge next to entries; reordering new cards by frequency relies on them);
- local pronunciation libraries (Japanese + English) — the speaker button in lookups, and the word audio on cards.
If all you do is watch anime and read in Japanese, the pack lasts a long time; there is no rush to add more.
Adding your own
Settings → Lookup → Dictionaries, import from file (multi-select is fine, import runs in the background), or drag .zip / .mdx / .dsl files straight onto the dictionaries page; the CSS that ships with an MDX dictionary must be imported together with the dictionary package. Imported dictionaries are grouped by category — monolingual / Japanese–English / bilingual / grammar / frequency / kanji / names / supplementary — and each can be:
- reordered: definitions appear in the popup in this order, so put the ones you read most first;
- hidden: ones you don't need for now aren't loaded and don't cost lookup time;
- collapsed: folded by default, open on tap.
Don't want to hunt for files? Download dictionaries in dictionary management lets you pick ready-made dictionaries by "Learning language" and "Your language" and download them in one go. Dictionaries that declare an online update URL (most Yomitan ones do) can all be refreshed with Update all dictionaries.
How many is enough? Yomitan power users run seventy or eighty, but every extra dictionary makes lookups a bit slower and the popup a bit longer. A set that gets you far:
| Category | How many | What for |
|---|---|---|
| Monolingual (J–J) | 1–2 | The workhorse. Learner-oriented small dictionaries have simple definitions and furigana; mid-size ones cover more words |
| Bilingual | 1 | Fallback when the monolingual entry loses you; use it less as you go |
| Pitch accent | 1 | High/low marks in the popup, needed for shadowing |
| Frequency | 1–2 | Deciding whether a word is worth a card, reordering new cards |
| Grammar | 1 | Look up patterns like 〜てならない without opening a grammar book |
| Kanji | 0–1 | Add when you want stroke order and on/kun readings |
Bilingual dictionaries and the commercial ones (物書堂 and friends) are copyrighted, so I won't list where to get them; the freely maintained Yomitan community dictionaries (the JMdict family, frequency lists, grammar dictionaries, Wikipedia summaries and so on) are enough to build this skeleton. Already have MDX dictionaries from your GoldenDict / Eudic days? Drag them in — no need to hunt for Yomitan versions.
Where pronunciation comes from
Settings → Lookup → Manage audio sources: several sources can be configured and are tried in order:
- Local audio database: the pronunciation library in the recommended pack is one of these; it is the same database format as Yomitan's "local audio server" (the NHK / Shinmeikai / JPod / Forvo set). If you already downloaded that data for Yomitan, "Add local audio database" pointed at its
entries.dbworks directly, optionally referencing the original file without copying. - Custom URL: any http(s) address that returns audio, with
{term}/{reading}placeholders. - Fushi interconnect: use the desktop's pronunciation library from your phone without installing the 9.5 GB, see interconnect.
With "Auto-play audio on lookup" on, the popup plays the first source's pronunciation as soon as it opens.
Common situations
- Unrecognised dictionary format: Yomitan dictionaries must be a zip (containing
index.json; old Yomichan ones are accepted too); for MDX select only the.mdx(plus the.mddnext to it) — don't unzip anything first. - MDX dictionary has no images / audio: the same-named
.mddwas left out; import it again. - Lookups got slow / popup too long: hide dictionaries you don't use, or collapse them.
- The same word repeated across several dictionaries: move bilingual ones to the end and collapse them; read the monolingual entry first.
- Fonts, colours, layout in the popup: "Dictionary styling" under Settings → Lookup, settable per dictionary; "Unify dictionary styles" (on by default) recolours the tags, emphasis and borders that imported dictionaries bring with them to the current theme's accent — turn it off to keep each dictionary's own look.