Type and image
Product-state truth
Labels and figures match from shot to shot, and an action shows its result.
Takeaways
- Same label, same figure, until something changes it.
- An action is followed by its result.
- Generated imagery supplies backgrounds, never the interface.
- Disclose a sped-up wait when it is material.
What it means
A viewer who sees 12 contacts in one shot and 14 in the next stops trusting every other number. The same goes for a button that is clicked and then nothing changes. Treat each shot as a state of one product: the figures persist unless something changes them, and every action is followed by the state it causes. Compressed demonstrations are fine when the cause and effect they show is true.
In a RasanAI film
UI words and numbers come from approved state data, not from a generative image, and are compared frame to frame. The order action, then state change is validated, and a sped-up wait is disclosed when it matters. Generated imagery may supply backgrounds, never interface text or controls.
Origins and sources
A production account of building interface text, clicks and highlights deterministically while AI supplied only backgrounds. The idea that a compressed demonstration must keep truthful causality is a RasanAI inference from it.
Evidence: Studio practice, and a RasanAI inference.