Korean AI voiceover

The Korean voices, with samples

Each sample reads the same line so you can compare them fairly.

Register is the first casting decision

Korean grammar wears its politeness on every verb, and the director sets it from your brief: haeyo-che friendliness for a cafe promo, hamnida-che formality for corporate work. Say which world your brand lives in and the script follows through consistently, narration and on-screen lines alike. The Language field's Korean entry locks ko-KR for the automatic voices, briefs in any language work, and quoted Korean in the brief, a menu item or slogan in Hangul, survives untouched into the video.

The eight ko-KR voices

All four style pairs exist here. Puck and Laomedeia serve fast retail and social spots, the register most Korean ad copy occupies anyway. Charon and Erinome read like a calm product tutor, and tutorials are locked to them by the standing rule. Umbriel and Sulafat handle testimonial warmth, right for a neighbourhood bakery or wedding studio. Iapetus and Aoede stay even for IR-style updates. One default voice per video, chosen by style and gender from the brief, with per-scene overrides and a full catalogue search available in each scene's voice dialog.

Konglish, numbers and respelling

English loanwords are everywhere in Korean commerce, and the voices read them with Korean phonology; when a brand must sound closer to its English original, respell it in Hangul that approximates the sound and regenerate, about half a credit per clip. Numbers need care because Korean counts in two systems: write the number as words in the intended system, 만 오천 원 for prices, and the ambiguity disappears. Sentence punctuation cuts the caption cues. Generated narration never carries an em dash by rule; keep manual edits on commas and full stops.

Timing behaviour and the bill

Korean's dense syllable blocks say a lot per second, so lines usually fit their scenes; when one overruns, the scene stretches to hold the clip and dynamic pace lets the tail play over the next cut. Each scene exposes a delay in seconds and speed steps from 0.85x to 1.25x. The soundtrack ducks itself to 22 percent while the voice speaks. Costs are the standard ones: narration billed by length, about half a credit for a typical clip, and an export that bills its render time and needs at least 1 credit left past 30 seconds or when narration remains to generate.

Captions in this script

Korean is the outlier among the CJK scripts because Hangul writes with spaces, so captions type on word by word in step with the voice exactly as English does, no segmenter required. Timing comes from the clip itself. The glyphs, though, come from the render host's fallback font, as no family in the catalogue ships Hangul; the syllable blocks render cleanly but in a typeface unrelated to the look's Latin. Korean particles keep words short, so the pill's 48 character sentence split rarely triggers; when a line does split, the cues cut between spaced words.

Questions people ask

Related pages