What it's good for
- Explainers, tutorials and courses where narration sits over screen recordings or graphics.
- YouTube channels adding audio tracks in other languages.
- Product and marketing videos for several markets.
- Animation, where characters' mouths are drawn loosely and timing can flex.
It voices the translated script you provide. It doesn't translate, transcribe your original or match lip movements, so on-camera talking heads need more care with timing.
How to localize a video
- Start from the original script with one line per shot or beat. If you only have the video, transcribe it first.
- Translate it, keeping the same line breaks, so line 12 in Spanish matches line 12 in English and its shot.
- Add the translation as a new script in the same project:
Launch video (German). - Add voices for the language. Language is a setting on each voice, so add a voice per language, from the original speaker's sample for a recognizable voice, or from a native speaker's for the most natural accent.
- Generate, review, save, then lay the numbered files against the same shots as the original.
Step-by-step detail is in translating a voiceover.
Timing: the hard part
Translations rarely run the same length as the original, and text translated from English into Spanish, French or German often comes out noticeably longer. Over a fixed picture that matters. Ways to handle it:
- Ask translators to fit the original's length, not just its meaning.
- Check a few long lines first, before generating everything.
- Give the edit a little slack: hold a shot longer, or trim pauses.
- Small tempo changes in the editor, under about 10 percent, are usually unnoticeable. Bigger gaps need a shorter line.
Accent and voice
A voice keeps the accent of the sample it was cloned from. Using the original presenter's sample keeps the video recognizably theirs, with some of their accent in the new language; a native speaker's sample sounds most natural. When using the original sample, lowering guidance strength in Fine-tune can reduce the carried-over accent. Each language page has writing tips for that language.
Review and publish
- Have a native speaker listen to every line before publishing. Automatic checks catch skipped words, not unnatural phrasing.
- Captions: the translated script comes in the download as text.
- Tell viewers the dubbed audio is AI-generated, and get the presenter's consent for their voice in other languages.