Korean, Japanese and Chinese subtitles that break correctly
Most subtitle tools treat every CJK script the same way and get two of the three wrong. Japanese and Chinese are written without spaces and break per character; Korean is written with spaces and must not be treated as spaceless. Here is what each script actually needs.
Start transcribing, freeThe mistake almost everything makes
Hangul sits in a Unicode range that most "is this CJK?" checks match, so tools written for Japanese and Chinese sweep Korean in with them. They then split Korean per character and rejoin with no spaces, and the output runs every word together: technically the right characters, unreadable as Korean.
Korean is space-delimited. It should be tokenised on whitespace exactly like English, and the spaces must survive being split and rejoined. That is a deliberate exclusion here rather than an accident of which range a check happened to cover.
Japanese and Chinese genuinely have no spaces
There is no whitespace to split on, so the unit has to be the character. That matches how forced alignment handles these scripts (alignment is per character) and it matches how the languages are actually typeset, where a line can break almost anywhere.
Rejoining then has to add no space between two of these characters, or こんにちは comes back as こ ん に ち は.
Latin text inside CJK
A Latin run inside Japanese (a product name, an acronym) is kept whole rather than split per letter, so a line break cannot land inside "AI". Where a Latin run meets CJK, a space is inserted, so "AI時代" comes back as "AI 時代".
That is one space more than a Japanese typesetter would use, and it is the deliberate choice. The rule requires both sides to be spaceless before it omits the space, because the alternative rule renders an English sentence containing one CJK word as "The時代of": merely untidy in one direction, unreadable in the other.
Reading speed is not the same either
The 42-characters-per-line guidance is for Latin scripts. CJK characters carry far more meaning each, so a line of the same character count takes considerably longer to read, and the usual advice is roughly half the character count per line. Check your longest cues rather than assuming a Latin-derived default transfers.
Translating into Korean or Japanese changes line lengths substantially. Timing boundaries are preserved through translation here, but a line that fitted in English may not fit after translation. It's worth a pass over the longest cues.
Frequently asked questions
Why do my Korean subtitles have no spaces between words?
The tool that produced them treated Korean as a spaceless script because Hangul is in a CJK Unicode block. Korean is written with spaces, and splitting it per character destroys them. Re-generating with a tool that distinguishes the two is the only real fix: the spaces cannot be reliably restored afterwards.
Why is my Japanese subtitle split in the middle of a word?
Japanese has no spaces, so a tool splitting on whitespace treats a whole sentence as one token and breaks it wherever the line runs out. Breaking per character is correct for the script.
Does word-level timing work for Japanese and Chinese?
Yes, per character, which is the meaningful unit in those scripts. Each character gets its own start and end time.
How many characters per line for CJK subtitles?
Considerably fewer than the roughly 42 used for Latin scripts: each character carries more, so the line takes longer to read. Judge it by reading time rather than by character count carried over from English.
Ready to try it?
Try it free without an account. Sign in to download your file, plus 15 minutes and 5 transcriptions a month.
Open the subtitle generatorNeed more than that? See pricing : one-time packs from $5, and nothing expires.