# Demo footage `ai_talk.mp4` is AI-generated video (a person at a laptop who does not exist), 6 s, 640×430, supplied by the maintainer for the demos. The `reframe_talk` demo uses it. It is not part of the npm package. `ai_talk_en.mp4` is AI-generated too (a man outdoors who does not exist), 15 s, 854×480, English speech with two hesitations. `ai_talk_en.words.json` holds its word timings as faster-whisper (base, primed with the filler list the way `silence.py --filler --transcribe` primes it) wrote them, so the `filler_en` and `pauses_en` demos, which both read this clip, build without a speech engine. These two clips are the only footage in `demos/build.py` that is not synthetic: three demos (`reframe_talk`, `filler_en`, `pauses_en`) use them, and every other demo is generated from FFmpeg's own test sources.