Problem
The assembly instruction for a Kinderkraft stroller had to reach customers in 18 languages, in landscape for YouTube and portrait for social media, with music, effects and a voice-over. By hand that is 36 separate edits and 36 places for a typo.
What I built
The script assembles each version from one source: a title card in the given language, the video with arrows and a step counter overlaid, a voice-over generated in ElevenLabs, music, effects and the claim at the end. The portrait version has its own framing and card layout.
Result
36 videos from one source, without a single manual editing session. A new language is a new row in the table and one run. A fix in the graphics or music lands in every version at once.
How it works
Card texts and voice-over lines live in one table, so the translator works in a spreadsheet, not in an editing program.
Screenshots and video
