Translating Korean web novels and Japanese Light Novels with AI: files and packs
Tri Thu知秋
· Updated
On this page
- How do Korean and Japanese novels count credits compared to Chinese ones?
- How do I keep furigana readings from doubling Japanese skill names?
- Are 제1화 and 第1話 chapter headings in TXT files split correctly?
- Which of the 5 glossary packs fits my Korean or Japanese novel?
- What are the 5 steps to translate a Korean or Japanese novel?
When switching from Chinese web novels to Korean web novels or Japanese Light Novels on NovelVerba, .epub files automatically strip furigana phonetic guides while plain .txt files do not. 제1화 and 第1話 chapter headings in a .txt file are recognized, but a few common variants such as 1話 without the leading 第, 외전, or 幕間 are not, so those chapters can end up merged. I prefer uploading an .epub file (or stripping furigana and fixing those few heading lines in the .txt file before uploading), then picking the right one of the 5 Korean and Japanese genre packs, so skill names never double and chapters stay cleanly separated.

| Source novel type on NovelVerba | Script & phonetic traits | Credit calculation | Recommended file format & matching Genre packs |
|---|---|---|---|
| Chinese web novels (ZH Simplified & Traditional) | Unspaced Hanzi lines, dense four-character idioms, and sect or clan seniority titles | Counts each character directly in the source text (including spaces): 1 character = 1 credit | EPUB or TXT (detects 第1章 automatically); full genre pack catalog on NovelVerba, including 5 packs built directly for Korean and Japanese novels |
| Korean web novels (KO Hangul) | Spaced Hangul syllable blocks, grammatical particles attached to nouns, and speaker-dependent honorifics | Counts each character directly in the source text (each Hangul syllable block like 헌 or 터 and each space = 1 credit) | Prefer EPUB. In TXT, 제1화 (even with a chapter title after it) and 1화 alone on its line are already recognized, so you only need to rename 1화 followed by a title to 제1화, move a title that contains punctuation (. , ! ? …) onto the next line, and replace 외전 with a standalone 番外; K-Novel: Hunter, Gate & Regression, Wuxia & Murim, Romance Fantasy & Villainess |
| Japanese Light Novels & web novels (JA) | Mix of Kanji, Hiragana, and Katakana with inline furigana readings over kanji or katakana skill names | Counts each character directly in the source text (including spaces); furigana readings in EPUB are stripped before counting | Prefer EPUB (auto-strips furigana readings and keeps the table of contents); for TXT, delete furigana annotations in your editor, rename 1話 without 第 to 第1話, and replace 幕間 / 閑話 with a standalone 番外 (第1話 is already recognized); J-Novel: Isekai & Fantasy, Romance Fantasy & Villainess, General: Infinite Flow & Unsure of Genre |
The first time I brought my Chinese novel habits over to a downloaded Japanese Light Novel .txt file, I opened Chapter 1 to find the holy sword エクスカリバー translated twice in a row because its phonetic furigana reading was sitting right next to the kanji. The chapter list was worse: the file numbered its chapters 1話, 2話... with no 第 in front, so all 80 chapters had been lumped into one giant chapter right underneath the opening プロローグ line. On the next Korean web novel, the 제1화, 제2화 chapters split correctly, but three 외전 side stories got glued onto the end of the main chapter before them. They looked like translation bugs at first, but they all came from how the .txt file was structured before I uploaded it.
How do Korean and Japanese novels count credits compared to Chinese ones?
When I drop a novel file into the upload box, the Source language field set to Detected automatically identifies Chinese (ZH), Japanese (JA), and Korean (KO) right away. For billing, NovelVerba applies the exact same rule across all three languages: it counts each character directly in the source text (including spaces), never billing by the length of the translated sentences. Each Hangul syllable block such as 헌 or 터, each Japanese Kanji or Kana character, and each Chinese character counts as 1 credit.
What changes is how the source files are written and formatted. Korean web novels put spaces between words and attach grammatical particles (like -이/가, -을/를, -께서) directly to noun stems, while Japanese Light Novels mix Kanji, Hiragana, and Katakana with phonetic furigana guides over skill names and proper nouns. To see how file preparation fits into the broader workflow for multi-hundred-chapter books, I recommend reading How to translate a whole web novel with AI: a 5-step workflow alongside Prepare your novel file before translating: pick a format and fix common issues.
How do I keep furigana readings from doubling Japanese skill names?
Japanese Light Novel authors frequently attach furigana phonetic guides above Kanji compounds to show a katakana reading or hidden meaning, such as writing 聖剣 (Holy Sword) with the reading エクスカリバー (Excalibur), or |エクスカリバー《約束された勝利の剣》. When you upload a Japanese .epub file, the app automatically strips the furigana readings above the kanji both when reading chapters and when exporting the translated EPUB for any target language other than Japanese. Only the base text remains, so furigana inside an .epub file never inflates your character count or duplicates skill names.
With plain .txt or .md files, however, the app does not automatically remove furigana (the plain-text furigana cleanup option is off by default and has no toggle on the novel creation screen). If I upload a .txt file from Syosetu, Kakuyomu, or Aozora that still contains inline furigana like |エクスカリバー《約束された勝利の剣》, |聖剣《エクスカリバー》, or 漢字(かんじ), the Kanji and Hiragana/Katakana sit right next to each other in the line. Every phonetic character adds to the source character count, and the AI ends up translating both parts side by side:
TXT file with inline furigana left in place: "勇者は|聖剣《エクスカリバー》を抜いた。" -> The Kanji and Katakana sit side by side, costing extra source characters and producing a doubled translation: "The hero drew the Holy Sword (Excalibur)." -> With an EPUB file (where furigana is stripped automatically) or after removing furigana in a text editor before uploading TXT: "The hero drew the Holy Sword."
Are 제1화 and 第1話 chapter headings in TXT files split correctly?
Unlike an .epub file, which is split into chapters along the content sections already packaged inside the book (usually one section per chapter), a .txt file (both when creating a novel and in the Convert a TXT novel to EPUB with a table of contents tool) is split by its heading lines. Recognized headings include: 第1章 and the same pattern with 話, 節, 回, or 卷 (第1話 or 第一話, even with a chapter title after it); Chapter 1; the Korean 제1화 (including 제1화 각성, as long as the line has no . , ! ? or …) and 1화 when it sits alone on its line; and standalone opening, closing, or extra lines such as 序章, 番外, プロローグ, エピローグ, 프롤로그, and 에필로그.
The variants that are not recognized are surprisingly common: 1話 without the leading 第, 1화 followed by a chapter title (1화 각성), a Korean heading with . , ! ? or … anywhere on the line (제1화. 각성, 제1화 각성!; colons, hyphens, and parentheses are still fine), an opening line with a subtitle (プロローグ 始まり), side stories and interludes like 외전, 幕間, 閑話, and 番外編, and Episode 1. A line that is not recognized never starts a new chapter, so everything under it gets glued onto the end of the chapter before it.
That is the プロローグ trap I ran into. When a .txt file has no recognized heading at all and runs past 600 lines, the app splits it into parts that aim for about 500 lines each. But a single standalone プロローグ line at the top is enough to switch that automatic split off, and if the chapters after it are numbered 1話, 2話... without 第, every one of them collapses into one massive chapter under that opening line. Fixing it in a text editor takes about 30 seconds. I only touch the lines that are not recognized and keep the chapter numbers as they are: change 1話 to 第1話, change 1화 각성 to 제1화 각성, move a subtitle onto the next line so プロローグ or 프롤로그 stands alone (a Korean chapter title with . , ! ? or … goes onto the next line too), and replace 외전, 幕間, 閑話, or 番外編 with 番外 on a line of its own. Never rename the opening line to Chapter 1, or every chapter number after it will shift by one.
Which of the 5 glossary packs fits my Korean or Japanese novel?
When creating a new project, NovelVerba's Genre menu offers 5 glossary packs tailored for Korean and Japanese fiction: K-Novel: Hunter, Gate & Regression (115 Korean Hangul entries), J-Novel: Isekai & Fantasy (120 Japanese entries), Romance Fantasy & Villainess (118 multilingual entries: 54 Korean + 34 Japanese + 30 Chinese), Wuxia & Murim (120 entries: 75 Chinese Wuxia + 45 Korean Hangul Murim), and General: Infinite Flow & Unsure of Genre (292 entries). The dictionary's substring matcher catches Korean noun stems automatically even when grammatical particles are attached to the end of the word (so 헌터들이 or 성좌께서 still match 헌터 and 성좌).
Because each novel project selects 1 genre pack when created, whenever I translate a hybrid story that blends two branches (such as a Korean Hunter who crosses into a Murim martial world), I select 1 main genre pack at creation time and add key terms from the secondary branch into My dictionary on the Shared dictionary page, or approve and lock them when running Scan names & terms. For deep dives into Korean and Japanese honorifics, pronouns, and genre terms, check out Korean honorifics in novel translation: keep oppa and hyung or translate them?, Korean web novel terms in translation: Hunters, Dungeons, Constellations, and Regression, and Japanese honorifics and pronouns in Light Novels: -san, senpai, Ore, and Boku. The term tables of the Wuxia & Murim, J-Novel: Isekai & Fantasy, and Romance Fantasy & Villainess packs each have an article too: Wuxia and Murim terms in translation: internal energy, sword qi, sword aura, and sect names, Isekai light novel terms in translation: reincarnation, transfer, the Adventurer's Guild, and magic, and Villainess novel terms in translation: noble ranks, engagement annulment, Condemnation Event.
| Genre pack in NovelVerba | Term count & source scripts | Best fit for Korean & Japanese novels and key terms |
|---|---|---|
| K-Novel: Hunter, Gate & Regression | 115 Korean Hangul base entries (KO) | Modern Korean web novels featuring Awakening (각성자 -> awakened), Red Gates (레드 게이트 -> Red Gate), Dungeon Breaks (던전 브레이크 -> dungeon break), Constellations (성좌 -> constellation), Regression (회귀 -> regression), Tower of Trials, and Chaebols |
| J-Novel: Isekai & Fantasy | 120 Japanese base entries (Katakana & Kanji JA) | Japanese Light Novels with Isekai reincarnation and summoning, Adventurer's Guilds, Demon Lords, Heroes, Unique Skills (ユニークスキル -> Unique Skill), and Magic Academies |
| Romance Fantasy & Villainess | 118 multilingual base entries (54 KO + 34 JA + 30 ZH) | Korean RoFan and Japanese Otome Villainess stories featuring the Grand Duke of the North (북부대공 -> Grand Duke of the North), Crown Princes, broken engagements, and condemnation events (断罪イベント -> condemnation event) |
| Wuxia & Murim | 120 base entries (75 Chinese Wuxia + 45 Korean Hangul Murim) | Korean Murim martial arts novels (화산파 -> Mount Hua Sect, 천마신교 -> Heavenly Demon Divine Cult, 구파일방 -> Nine Great Sects and One Gang) and classic Wuxia |
| General: Infinite Flow & Unsure of Genre | 292 base entries (covering core system, dungeon, and transmigration terms) | System, LitRPG, survival, or multi-genre hybrid web novels when you want broad coverage across status screens, dungeons, and transmigration |
What are the 5 steps to translate a Korean or Japanese novel?
- Prepare your source file (.epub or .txt): Prefer an .epub file so the app automatically strips furigana readings and preserves the table of contents. If starting from a .txt file, I remove furigana annotations in my text editor first so phonetic characters do not waste credits or double skill names, then fix the heading lines that are not recognized: 1話 becomes 第1話, 1화 followed by a title becomes 제1화 with the same title (if the title has . , ! ? or …, put it on the next line), and 외전 / 幕間 / 閑話 become a standalone 番外, before uploading or using TXT → EPUB with a table of contents.
- Create the novel and pick 1 of the 5 Genre packs: Upload your file, leave Source language on Detected automatically, choose your target language, and select K-Novel: Hunter, Gate & Regression, J-Novel: Isekai & Fantasy, Romance Fantasy & Villainess, Wuxia & Murim, or General: Infinite Flow & Unsure of Genre; add terms from any secondary genre into My dictionary on the Shared dictionary page.
- Run Scan names & terms on the first 5 to 10 chapters: Click Scan names & terms so the AI extracts character names, Constellation modifiers, and custom skill names, add pronoun or honorific notes (such as Ore, Boku, Watashi or oppa, hyung) into the Note field while reviewing the scan results, and lock them before translating.
- Set up Addressing by character and Style pairs: Turn on Character speech mode (choose AI guesses or Main character for first-person Light Novels) and pin 3 to 5 dialogue examples in the Style tab.
- Translate chapter 1 and verify bilingually: Translate the opening 1 to 3 chapters first, check skill names and dialogue voice in the bilingual reader, and then translate the rest of the novel in batches.