interview-reel-motion

Make an interview short read clearly with one graphic message at a time, never over a face, and never in the platform UI. You write a beat map; the template renders it; the QA scripts prove it.

The default engine is HyperFrames (template/): build.py turns three JSON files into a page (frame stacks, audio mix, the scene), runtime/ holds the engine and the kit, npx hyperframes render (or lambda render) encodes it. scripts/ holds the QA gates and the pre-passes, tools/ramp.py builds speed ramps. No app needed: Node 22+ (the HyperFrames CLI requires it), Python 3 with numpy and pillow, ffmpeg and ffprobe on PATH (or set IRM_FFMPEG and IRM_FFPROBE). The QA gates add pip install -r scripts/requirements.txt and the MediaPipe face_landmarker.task model (references/qa.md).

When to use

• The short is already cut (clip in and out points, voice) and needs its graphic layer, cover and ending.
• Any machine with Node, Python and ffmpeg. The QA scripts work on any MP4.

Not for: cutting phone footage into a short (vlog-edit does the cut and its own cards), narrated product films, or a motion graphic with no speaker.