Children's captions at 17 characters per second: Netflix rule and Sume

What the Netflix guide says

The Netflix English Timed Text Style Guide gives a reading speed of 20 characters per second for adult programs and 17 for children's programs. It also sets 42 characters per line and a maximum of two lines.

Reading speed is the number that decides whether a caption feels rushed. A line of 34 characters at 17 characters per second needs two full seconds on screen.

What Sume lets you control

Sume's caption design.phrasing group has max_words, max_chars and pause_seconds. These shape how spoken words are grouped into on-screen phrases. The docs describe them as grouping controls; no reading-speed setting is documented, and cards follow the speech timing.

What you can do is make cards smaller, so fewer characters need to be read in the time the speaker takes. A card with fewer characters at the same duration has a lower reading speed.

A practical approach for children's content

Burned-in captions on children's video can be tuned toward the slower guide without a reading-speed field.

• Lower max_chars so each card carries a short phrase.
• Use pause_seconds so cards break at natural pauses.
• Raise font size for small-screen viewing with typography.font_size_ratio.
• Slow the voice. If you generate the speech with Sume TTS, generation_config.speed accepts 0.6 to 1.5, so a slower read also lowers characters per second.

Check the numbers yourself

Take the cue list, compute characters divided by duration for each, and flag any above 17. A short script is enough. Where a card exceeds the limit, shorten the wording or split it.

Remember that Netflix's numbers are for its own delivery; if your destination has no such rule, this is guidance, not a requirement. See the 42-character rule post for line length.

Measuring your cues

A short script settles it. For each cue compute characters divided by seconds on screen, and list those above 17. Count characters the way the guide implies, as the visible text including spaces, and treat the guide itself as the authority for edge cases.

If many cues fail, the voice is probably too fast for the audience. Slowing the narration is better than shrinking the text, because it helps listeners as well as readers.

Takeaway

Use 17 cps as a target, shorten cards via phrasing, and slow the narration if you made it. Sume does not enforce reading speed, so verify it with the timings you have.