BEHIND THE CAPTIONS
Readable by design.
The studies that inform our captions, the choices we made, and the questions we’re still testing.
These are starting points, not proof of a retention boost. Ad research and prerecorded subtitle studies do not directly validate live Twitch captions.
TikTok's creative research supports contextual text and selective emphasis. It studies advertising—not long-form Twitch caption comprehension.
In AuraMax Short phrases and one accent at a time. No constant bouncing.
TikTok / Kantar, 2022 ↗At 12, 20, and 28 characters per second, faster subtitles produced less complete reading and lower comprehension. Other studies found viewers could cope with 20 cps; context matters.
In AuraMax Adjustable phrase hold and stable lines, rather than a universal speed claim.
Kruger, Wisniewska & Liao, 2022 ↗A small study of Chinese viewers found dual subtitles did not create greater cognitive load or greater gains than monolingual subtitles. It does not prove bilingual is best for everyone.
In AuraMax Offer original, translation, or both. Keep translation placement consistent.
Liao, Kruger & Doherty, 2020 ↗TikTok recommends keeping key content inside UI safe zones and testing creative variations. Its vertical-ad layout is not a Twitch safe-area specification.
In AuraMax Preview both formats, move captions, and inspect them against actual footage.
TikTok Creative Best Practices, 2025 ↗STILL TO LEARN
We still need to test comprehension, fatigue, timing and preferences with English and Chinese viewers over real Twitch footage. Animation durations and phrase timing are product choices, not scientifically established optima.
A contrasting 2018 subtitle-speed study found viewers could cope with fast subtitles. Context and soundtrack familiarity matter.
See the audience statistics ↗