How Fast Do People Speak on YouTube? Japanese and English Speaking Speed by Level

Japanese YouTube videos are spoken at a median 6.3 morae per second and English videos at 183 words per minute. Beginner Japanese (JLPT N5) videos run 36% slower than N1 videos.

Key findings

• Japanese: median 6.3 morae per second across 1,342 videos.
• JLPT N5 videos: 4.4 morae per second; N1 videos: 6.9.
• English: median 183 words per minute across 1,300 videos.
• CEFR A1 videos: 144 words per minute; B2 videos: 198 (27% slower at A1).

Japanese speaking speed by JLPT level

Speed is counted in morae, the rhythm units of Japanese (たべる is three: た・べ・る; きょう is two: きょ・う). Across 1,342 videos the median is 6.3 morae per second. Speed climbs steadily from N5 to N3 and then flattens: N3, N2 and N1 videos differ far more in vocabulary than in pace.

If native Japanese feels too fast, why native Japanese sounds so fast covers the reasons beyond raw speed and how to train for them.

English speaking speed by CEFR level

Across 1,300 English videos the median is 183 words per minute. A1 videos (mostly made for learners and children) are clearly slower; from B1 up, speed sits close to ordinary conversation, and C1 videos are not faster than B2 because they are harder in content, not pace.

Fastest and slowest channels

Channel style matters as much as level: the fastest Japanese channel here (8.4 morae per second) speaks about twice as fast as the slowest (4.0), and the English range runs from 227 down to 91 words per minute. Only channels with at least 10 measured videos are listed, fastest first.

Method

• Corpus: 2,518 Japanese videos (2,110,066 words, 233 hours of speech) and 2,146 English videos (2,447,824 words, 229 hours of speech) from YouTube creators, as processed by LingoDew up to 2026-10-16. Only videos with at least 60 seconds of speech are measured (short clips give unstable rates), which leaves 1,342 Japanese and 1,300 English videos.
• Speaking time is the sum of subtitle sentence durations from LingoDew's word-level audio alignment: silence between sentences is excluded, short pauses inside a sentence are included.
• Japanese morae are counted from the kana reading of every word (small ゃ・ゅ・ょ join the kana before them; っ and ー count as one mora each). 100.0% of words have a reading; the rest (mostly numbers and Latin letters) are not counted, so the Japanese figures lean slightly low.
• English words are the words in the subtitles, so a contraction like don't counts as one word.
• Each video gets one rate; tables show the median video, so a few very fast or slow videos do not move the result.
• Levels are the difficulty labels LingoDew shows on every video (JLPT N5–N1 for Japanese, CEFR A1–C2 for English). They are estimated from vocabulary, grammar and speed, and some are reviewed by hand. Levels with fewer than 20 videos are left out.

Download the data

• Speed by level and channel (CSV)
• This report as plain text (Markdown)

Text, tables and data are free to reuse underCC BY 4.0: credit “LingoDew” and link to this page.