Add Japanese Subtitles to a Video

Text without spaces

In English, a subtitle line can break at any space. Japanese gives no such hint: 今日は新しいカメラを試してみます is one unbroken run of kanji and kana. Break it in the wrong place and a word is cut in half, or a closing bracket lands alone at the start of the second line.

The editor deals with this by asking the browser's built-in word segmenter where the words are. Inside a Japanese run, those word boundaries become the only places a line may wrap, so 新しい and カメラ each stay whole. The same segmenter handles Chinese and Thai; Korean is different because it writes spaces.

Two consequences follow. A line breaks where the words fall, which is usually close to a natural pause but not always where a professional subtitler would choose. And when you want a specific break, the reliable way is to type it yourself: press Enter inside the line's text box, and the editor keeps your break and wraps each part separately.

The kinsoku rules it follows

Japanese typesetting has long-standing rules, called kinsoku shori, about characters that may not start or end a line. The editor applies the basic set when it wraps:

• Never at the start of a line: closing punctuation and brackets (、 。 」 』 ） ！ ？ ： ；), the long-vowel mark ー, the repetition mark 々, and the small kana ぁ ぃ ぅ ぇ ぉ っ ゃ ゅ ょ ゎ and their katakana forms such as ッ ャ ュ ョ.
• Never at the end of a line: opening brackets 「 『 （ 【 〔 〈 《.
• Characters in those groups are glued to their neighbour before the line is measured, so 「です。」 travels as a unit rather than leaving 」 stranded.

Once the break points are known, the editor balances the lines: if a subtitle needs two lines, it narrows the width until both are close in length instead of leaving one word on the second line. Only a single unit wider than the whole line, such as a long URL, is ever cut between characters.

Sixteen characters, eight per second

Japanese packs more meaning into each character than English does, so the limits are lower. The editor flags a Japanese line longer than 16 characters (it still wraps) and a subtitle that asks for more than 8 characters per second. Kana, kanji and full-width punctuation all count as one character each; spaces don't count toward reading speed.

The language of a line is detected from its characters, and any hiragana or katakana marks it as Japanese. A line made only of kanji, such as 東京大学, carries no kana, so it is checked against the Chinese limits (9 per second) instead. That rarely matters, but it explains the occasional different warning.

AI output is built around those numbers. Lines come back at about 12 characters, short enough for one line on a phone, each shown for up to about 2.6 seconds. In practice, when a line still reads as too dense:

• Cut sentence-final padding that the speech had but the reader doesn't need, and keep the core verb.
• Prefer one clear clause per subtitle; a second clause can go in the next line.
• Hold a dense line a little longer by dragging its end edge, as long as it doesn't run into the next one.

Subtitle segmentation covers where a sentence should be divided, whatever the language.

Getting the Japanese text

• Japanese speech: leave "Subtitle language" on auto-detect. Whisper transcribes the Japanese with word timings, so each line starts on its first spoken word.
• Speech in English or another supported language: set "Subtitle language" to Japanese. The lines are translated, and their timing is shared out by text length, so check that each one starts with its phrase. Decide whether the video should speak in polite です/ます forms or plain forms, and make the lines consistent; burning translated subtitles has a review routine for translated lines.
• A subtitle file from a translator: switch AI off and use "Import SRT / VTT".
• Your own typing: each line's text is an ordinary text box, so your Japanese input method works as usual, including Enter to confirm a conversion.

AI subtitles cost 1 credit per minute, so a 25-minute vlog is 25 credits (2.5¢); 100 free credits come with a new Google sign-in. Accuracy is lowest on names, place names written with unusual kanji, and speech over music, so read every line before rendering.

Which font draws the kana and kanji

None of the 19 fonts in the Style panel contains Japanese characters. The editor therefore falls back to Noto Sans JP for kana and kanji, loading it only when Japanese text is present, while letters, digits and Latin punctuation keep the font you picked. If the web font can't be fetched, installed fonts such as Hiragino Sans or Yu Gothic take over.

That has practical effects on styling:

• Pick a Latin face that sits well beside a clean Japanese gothic. Noto Sans, Inter and Roboto match its proportions; Bangers or Playfair Display will look like a different design in the middle of a sentence.
• The Japanese fallback is loaded in regular and bold, so finer weight steps change the Latin letters but not the kana.
• Italic and UPPERCASE have no effect on kana and kanji, which have no slanted or capital forms.

For outline and box choices that hold up behind dense characters, see customizing subtitles.

What it won't do for Japanese

• Vertical text. Lines are drawn horizontally only; tategaki subtitles, sometimes used for side text in Japanese film, aren't available.
• Furigana. There is no ruby text above kanji. If learners need a reading, add it in parentheses after the word, or choose a kana spelling.
• A clean Split. On a line with no spaces, Split divides the text at the character matching the playhead's position, which can land inside a word. Check both halves afterwards and move a character or two by hand.
• Speaker names. The AI doesn't say who is speaking; add a name in brackets yourself where a scene needs it.

Two lines worked through

Checklist before rendering

• Every line read against the speech, with names and numbers checked.
• No line flagged for reading speed that you haven't either shortened or lengthened.
• Breaks where you want them; type Enter inside the text for any you want to force.
• A Latin font that matches Noto Sans JP, previewed on a bright and a dark scene.
• Text clear of on-screen captions in the original footage, using a vertical offset if needed.

For a Japanese SRT without a new video, the Japanese subtitle generator is the shorter route, and Japanese video to text produces a transcript. Going the other way, from Japanese speech to English viewers, the Japanese to English video translator covers subtitles and dubbing. Chinese follows similar line rules with its own limits, set out on adding Chinese subtitles.

About the subtitle editor on this page

Opens

MP4, MOV, M4V, WebM, MKV videos your browser can decode

Subtitles from

AI (from the speech, optionally translated into one of 21 languages), an SRT or VTT file, or typing them in

You download

An MP4 with the subtitles drawn into the picture at the original resolution and frame rate, plus SRT and VTT files

Cost

Editing, styling, rendering and SRT/VTT export are free, with no account and no watermark. AI subtitles cost 1 credit per minute (5-credit minimum per video); signing in with Google gives 100 free credits and accounts get 10 free credits a day

Privacy

The video never leaves your device and is rendered in your browser. Only for AI subtitles is the audio sent, over HTTPS, and it is deleted within 30 minutes of delivery

Rendering needs

Chrome or Edge 94+, Safari 16.4+, or Firefox 130+ on a computer. Editing and SRT/VTT export work in any current browser