How Long Does a Mandarin Line Take? A Characters-to-Seconds Table for Short Drama
Credits and specs follow TWno1's live billing settings. Last checked: 2026-10-05
Plan Mandarin dialogue at about 4 characters per second: a 16-character line takes roughly 4 seconds to say.
That figure sits between two published measurements, one fast and one slow. Below is a lookup table, then three things the table alone will not tell you. Subtitle reading speed rarely limits a dubbed drama. A 6-second shot cap does. And many video models start at 3-4 seconds a clip, so short lines need trimming.
The lookup table
| Characters | Slow (3.8/s) | Typical (4/s) | Fast (5.2/s) | Suggested shot (4/s + 0.5 s) | Subtitle lines (16 per line) |
|---|---|---|---|---|---|
| 4 | 1.1 s | 1.0 s | 0.8 s | 1.5 s | 1 |
| 8 | 2.1 s | 2.0 s | 1.5 s | 2.5 s | 1 |
| 12 | 3.2 s | 3.0 s | 2.3 s | 3.5 s | 1 |
| 16 | 4.2 s | 4.0 s | 3.1 s | 4.5 s | 1 |
| 22 | 5.8 s | 5.5 s | 4.2 s | 6.0 s | 2 |
| 25 | 6.6 s | 6.3 s | 4.8 s | 6.8 s | 2 |
| 32 | 8.4 s | 8.0 s | 6.2 s | 8.5 s | 2, split into two subtitles |
| 40 | 10.5 s | 10.0 s | 7.7 s | 10.5 s | 3, must split |
The speed columns are plain division: characters ÷ characters per second. The shot column adds 0.5 s. Most Mandarin characters are one syllable, so syllable rates work as character rates. Punctuation costs time but no characters, which is one reason we plan on the slower side.
Where 3.8, 4 and 5.2 come from
5.18 syllables per second is from Pellegrino, Coupé and Marsico, published in *Language* in 2011. Speakers read the same texts aloud. Mandarin came in at 5.18 syllables/s, English 6.19, French 7.18, Spanish 7.82, Japanese 7.84. Mandarin was the slowest of the five. Read-aloud speech has a script and no hesitation, so treat it as the fast end.
3.8-4.1 characters per second is from Yuan, Liberman and Cieri at Interspeech 2006. They measured telephone conversation corpora: about 247 characters per minute in CallHome/CallFriend and 228 in Fisher. Divide by 60 and you get 3.8 to 4.1. That is real talk, with pauses, breaths and fillers included.
Scripted drama lands in between. Actors have lines, but they also act, and acting includes pauses. We default to 4 per second and push toward 5 for arguments and overlapping dialogue.
One more figure gets misused. A 2019 study in *Brain and Behavior* (30 participants) measured a maximum Chinese reading speed of 259.5 characters per minute, about 4.3 per second, with words flashed one at a time. That is a lab ceiling for single-line flashing text. It is not a comfortable subtitle speed, so do not build subtitle rules on it.
Subtitle speed is almost never the bottleneck
Netflix's Traditional and Simplified Chinese style guides set 16 characters per line, 2 lines maximum, and up to 9 characters per second for general programming. Each subtitle event stays on screen for at least 5/6 of a second (20 frames at 24 fps) and at most 7 seconds.
Compare 9 per second with speech. A 16-character line takes about 4 s to say and 16 ÷ 9 ≈ 1.8 s to read. When subtitles follow the voice, viewers finish reading long before the actor finishes talking. In a voiced drama the reading cap rarely binds.
Two other limits do bite:
- The 7-second cap. Two full lines is 32 characters. At 4 per second that is 8 seconds of speech, over the 7-second maximum. A line that fills both rows needs two subtitle events.
- 16 characters per line. Anything from 17 characters up wraps to a second line, and two lines cover a lot of a phone screen.
As one formula: subtitle duration = max(spoken time, characters ÷ 9, 0.83 s), then cap at 7 s.
English works differently, so do not borrow its rules. Netflix's English guide allows 42 characters per line and 20 characters per second, numbers that make no sense for Chinese.
The real ceiling is the 6-second shot
On our demo drama we capped single shots at 6 seconds for pacing and sized each shot as line length + 0.5 s. Put those together: (6 − 0.5) × 4 = 22 characters.
So at a normal pace one shot holds about 22 characters of dialogue. In a 16-episode demo series we reviewed, lines averaged 12.7 characters and the longest was 22.
The script rules we used on that demo drama:
- Keep a line at 25 characters or fewer. Note that at 4 per second, 25 characters needs a 6.8 s shot, slightly over the cap.
- Flag anything at 40 characters or more as too long. That is about 10 seconds of speech. The fix is a longer clip or a shorter line.
Why AI clips often need trimming
Video models have a minimum clip length. Google's Veo 3.1 offers only 4, 6 or 8 seconds, and Kling 3.0 runs 3-15 seconds, per their official docs checked 2026-10-05. Some go shorter: Runway Gen-4.5 starts at 2 seconds, Vidu Q3 and Grok Imagine at 1. A three-character line like 你來了 ("you're here") takes under a second to say. On Veo the clip starts at 4 seconds. Left untouched, the scene plays like slow motion.
So trim to line length + 0.5 s, within ±0.5 s. The rule that matters is the other direction: never cut earlier than 0.5 s after the line ends. We learned this on the demo series. One trimmed shot lost the final syllable of its line. Another opened with 1.9 seconds of silence, and its subtitle sat on the character's chest. On the same 7 shots, the old trimming met the length target 0 times; trimming to line + 0.5 s met it 7 times.
Voice speed is the other trap. The demo series switched narration voices once, and the new voice read 70% slower than the old one. The user flagged it as dragging right away. The lookup table is for writing. Once the voice track exists, measure its real length and adjust the speed or the line.
Limits of this table
- These are averages. Crying, shouting and deliberate pauses move far from 4 per second. Use the table to size scripts; trust the measured voice track for the final cut.
- Mandarin only. In Pellegrino's data, Japanese and Spanish run above 7.8 syllables per second and English at 6.19. Using Mandarin rates for other languages will be badly off. We have no reliable figures for Cantonese or Taiwanese Hokkien.
- The studies are not new. 2011 and 2006. Speech rate changes slowly, but the conversation data is phone calls, not performance.
- One subtitle authority. We found no separate public Taiwanese subtitle standard. Netflix's guides are the most complete public rules we could find.
Catch long lines before they reach the storyboard. TWno1 takes a script you paste in and breaks each scene into a shot list you can edit before any frame is drawn. That is the place to lengthen the shot for a 25-plus-character line, or cut the line down, instead of finding out after the voice track is recorded.
Related reading: Vertical Drama Shot Length: Why We Aim for 2.5-4 Seconds and Cap Shots at 6; Scene Plates for AI Drama: Empty Master Shots, 180-Degree Reverse Views and Lighting Variants; Prop Continuity in AI Drama: Safes That Morph, Newspapers That Heal, Clocks That Lie.
FAQ
How many characters per second do Mandarin speakers say?
About 4 characters per second. Pellegrino and colleagues (2011) measured 5.18 syllables per second for read-aloud Mandarin, a fast upper bound. Yuan and colleagues (2006) found 228-247 characters per minute in telephone conversation, or 3.8-4.1 per second. Scripted drama sits between the two.
How long does a 20-character Mandarin line take?
A 20-character Mandarin line takes about 5 seconds at 4 characters per second, about 3.8 seconds at a fast 5.2 per second and about 5.3 seconds at a slow 3.8. For video, size the shot at line length plus 0.5 seconds, so roughly 5.5 seconds.
What is the maximum line length for Mandarin short-drama dialogue?
We recommend 25 characters or fewer per line and treat 40 or more as too long, fixed by a longer clip or a shorter line. As a second reference: with a 6-second shot cap and 0.5 seconds of tail, one shot fits about 22 characters at 4 per second. In our 16-episode demo series, lines averaged 12.7 characters and the longest was 22.
What is Netflix's reading speed limit for Chinese subtitles?
Netflix's Chinese style guides allow up to 9 characters per second for general programming, 16 characters per line and 2 lines, with each subtitle shown between 5/6 of a second and 7 seconds. Because Mandarin speech runs near 4 characters per second, the 7-second cap is the limit dubbed dramas usually hit, not the 9 per second.
How much should I trim an AI video clip that is longer than the line?
Trim the clip to the line's spoken length plus 0.5 seconds, within ±0.5 seconds, and keep shots without dialogue at 2-3 seconds. Never cut sooner than 0.5 seconds after the line ends, or you may clip the last syllable. Veo 3.1 only offers 4, 6 or 8 second clips.
References
- Pellegrino, Coupé & Marsico (2011), *Language*: cross-language read-aloud speech rate (accepted manuscript, PDF) (checked 2026-10-05)
- Yuan, Liberman & Cieri (2006), Interspeech: speaking rate in telephone conversation corpora (PDF) (checked 2026-10-05)
- Netflix Traditional Chinese Timed Text Style Guide (checked 2026-10-05)
- Netflix general timed text requirements (checked 2026-10-05)
- Netflix English Timed Text Style Guide (checked 2026-10-05)
- Maximum Chinese reading speed under rapid serial presentation, Brain and Behavior (2019) (checked 2026-10-05)
- Google Gemini API: Veo video generation (checked 2026-10-05)
- Kling 3.0 user guide (checked 2026-10-05)