GetBereit TikTok

Short vertical videos that teach German for the ÖSD and ÖIF exams, built in Remotion and ending in a call to download GetBereit. The teaching is the ad — nobody watches a promo for an exam-prep app, but they will watch a thirty-second explanation of the accusative case.

Not a template engine

The obvious build is a set of templates with slots to fill. This is the opposite: an authoring kit for a language model. Each video is its own bespoke component in src/videos/, hand-written for that idea, with no forced format and no schema to satisfy. If an idea needs something the kit has never done, it just gets written.

What keeps quality from drifting is a small shared kit rather than a rigid format: voice-synced kinetic captions, an audio-driven timeline, reveal hooks, and the CTA.

The one thing that is standardised

Narration. A video declares the lines it speaks, and the pipeline generates the voice and the per-word timings before rendering a single frame. Everything visual can then key off the audio instead of a timer.

That is the difference between video that feels alive and video that feels generated. Captions pop one to three words at a time, in sync with the voice. Elements enter on the exact frame a word is spoken — when the narrator says Akkusativ, that is the frame the card lands on, because wordCue says so. No dumb timers, and nothing to re-time by hand when a line is rewritten.

Voice is Azure Neural TTS in Austrian German, which matters for an Austrian exam.