Long Japanese recordings to text, in natural Japanese script
Upload a Japanese audio or video file of up to 6 hours and get the full transcript in ordinary Japanese writing, kanji, hiragana and katakana together, with a timestamp at the start of every paragraph. Add an English translation underneath if a colleague or supervisor needs it. Up to 15 minutes is free in full; on a longer recording the first part is free and the whole file is $5, paid once.
- Kanji, hiragana and katakana mixed as in normal Japanese text, with no spaces between words
- Keigo kept as spoken: desu and masu forms, honorific and humble verbs are not flattened
- English loanwords written in katakana when they are said as Japanese words
- Romaji and furigana are not offered
- Up to 6 hours per recording for $5; 12 hours with Unlimited
Real output from a demo recording. Click a timestamp to see where it sits in the audio.
How Japanese speech is written here
Japanese is written with three scripts at once and without spaces between words. The transcript follows ordinary written conventions: kanji for most content words, hiragana for grammar and many everyday words, katakana for loanwords such as ミーティング or プロジェクト. Each paragraph starts with a timestamp in Western digits, like [1:05:12], followed by the Japanese text.
The choice between kanji and kana is not fixed in Japanese; the same word can be written either way by different writers. If your style guide prefers one form, use Find and replace on screen to make it consistent before you download.
Keigo, casual speech and Kansai-ben
Speech level carries meaning in Japanese, so the tool writes it as spoken. A client meeting keeps its いたします and おっしゃる, a chat between friends keeps its plain forms, and nothing is converted into the other. That matters when you analyze an interview: how politely someone speaks is part of the data.
Regional speech is written as heard as far as the model can follow it. Kansai expressions such as あかん or ほんま stay in the transcript instead of being swapped for standard Japanese. Strong dialect from Tohoku, Kyushu or Okinawa is harder and more likely to contain mistakes; check the free first part.
Names, numbers and other things to check
- Personal and place names: one reading can be written with many different kanji, and the transcript has to pick one, so confirm names against a list or a business card
- Numbers may appear as digits or in kanji; make them consistent if it matters
- Speaker labels: numbers are assigned per block of about ten minutes, so a long seminar may give the same person different numbers
- English terms said as English, not as loanwords, may appear in Latin letters
Who brings long Japanese recordings
- Researchers with fieldwork interviews in Japan, needing the Japanese plus an English version
- Students with seminar recordings or interviews for a graduation thesis (卒業論文)
- Foreign staff at Japanese companies who need meetings in writing with English underneath
- Families recording grandparents, for grandchildren who read English better than Japanese
How the tool works, what it costs and how your files are handled is on the Long Audio & Video to Text page; prices for every tool are on the pricing page.
Last updated
Questions and answers
Can ChatGPT transcribe a two-hour Japanese meeting?
Chat assistants can manage a short clip. A long meeting tends to exceed their audio limits or come back shortened, without dependable timestamps. This tool handles the recording in parts of up to ten minutes and puts the full text back together in order.
Can I get romaji or furigana?
No. The transcript is written in standard Japanese script. Choose English under "Also translate" if the reader does not read Japanese.
Is keigo preserved?
Yes. Polite, honorific, humble and plain forms are written as they were spoken and are not normalized.
What does a long seminar cost?
One payment of $5 covers a recording of up to 6 hours. A file of up to 15 minutes is free in full. Unlimited, $19 a week, covers up to 12 hours per job; cancel any time.
Is the recording kept private?
Only you can open it. The audio goes to an AI model provider purely for transcription and is not used for model training. Jobs created without an account are deleted after 7 days, or earlier if you delete them.