Romanization guidelines
On this page
This guide assumes the reader has basic knowledge of Japanese grammar to be able to correctly romanize.
See also the wiki page for Romanization walkthrough.
Follow these romanization guidelines when romanizing titles.
For Japanese, the rules are based on the Hepburn system with some exceptions.
For Chinese, use the Pinyin system.
Do not add machine translated/romanized titles or lyrics.
Official translations are accepted in any case.
If you are unsure about the romanization, leave a comment on the entry. You can use @(username) to notify the editor who added the romanization.
Voice synth-related Japanese slang and terminology is collected in this Google Spreadsheet.
The official romanization and/or translation should be prioritized even if it’s against VocaDB’s romanization guidelines.
New non-English title translations should not be added without a good reason.
For example, the song entry ECHO should not receive new name/title translations.
Acceptable exceptions:
- Translation is used somewhere elsewhere (add for improved searchability).
- Translation is related to the original name (additional context/value).
For example, マトリョシカ is from the Russian word “Матрёшка”, but is also known by its Romanized form “Matryoshka”.
Separate all words and particles with spaces when romanizing.
For Pinyin, separate each character with a space when romanizing.
Special names and proper nouns are kept together.
Capitalize all words that are NOT single kana particles (は、が、を、etc).
- E.g. 歌に形はないけれど would be romanized as ‘Uta ni Katachi wa Nai Keredo’.
Do not use macrons for long vowels.
- こおり is romanized as ‘koori’, not ‘kōri’.
If ん is followed by a vowel (including や・ゆ・よ), romanize it as n’.
- E.g. してやんよ would be romanized as “Shiteyan’yo’“.
は (ha) is romanized as ‘wa’ if it is used after a noun as a topic marker.
- E.g. 恋は戦争 would be romanized as ‘Koi wa Sensou’, the は is not part of a word and functions as a particle.
を (wo) is romanized as ‘o’ if it is used after a noun as a direct object particle.
- E.g. 地球最後の告白を would be romanized as ‘Chikyuu Saigo no Kokuhaku o’.
へ (he) is romanized as ‘e’ if it is used after the name of a place to indicate direction as a particle.
- E.g. 駆ける、星空の彼方へ would be romanized as ‘Kakeru, Hoshizora no Kanata e’, the へ is not part of a word and functions as a particle.
Use the same order as the original Japanese name for romanizations.
Usually this is “Lastname Firstname”.
Loanwords are romanized into their original language.
- コーヒー would be romanized as ‘coffee’.
Use a dash with honorific suffixes on names when romanizing.
- お姉さん would be romanized as ‘Onee-san’.
If the entire name is composed of loanwords, leave the romanized name field empty. Use the translated name field instead.
First of all, thank you for your help!
Most of the above guidelines need not apply for lyric romanizations, but we request that you do follow the guideline of not using macrons for long vowels. Otherwise, it is up to your discretion.
Always leave a source for the lyrics. If transcribed/translated by yourself, mention that instead.
If the lyric source is missing, locate it before modifying the lyrics. See a list of websites where to look for lyrics. If the source can’t be found, create an entry report for the missing source and use a credit like “Unknown (fixed by user2)“.
Many older songs (and some newer songs) do not provide lyrics in their descriptions or elsewhere, and their lyrics can only be found in the PV itself. A decent amount of other songs do not have lyrics in the PV either, and therefore their lyrics have to be transcribed by hearing.
Visual transcriptions are almost always easier than auditory ones, as you can rely on both seeing the characters themselves and having the readings to go with them from the song’s audio.
- If you transcribe from a PV, make sure to link the PV as the source and add (Transcribed from PV) next to it. For example, “YouTube (Transcribed from PV)“
- The use of OCR (Optical Character Recognition), such as Google Lens and other AI-assisted tools, is allowed, but they should not be blindly trusted. If you use them, make sure to confirm that they are accurate before adding them to an entry. Keep in mind that OCR is often quite unreliable on stylized or low-quality text.
Audio transcription is much more difficult than visual transcription, duing to the wide variety of characters that share similar or identical readings. Transcribers need to not only know these characters, but also decide which one fits the given context best. For this reason, it is generally discouraged to attempt audio transcriptions unless you are advanced enough in the languge.
Ai-assisted tools are not allowed for audio transcription, as they are extremely inaccurate with languages like Japanese and Chinese and even more so with Vocaloid.