NotebookLM's cinematic video output draws complaints

This reads less like a single bad prompt and more like a rendering or presentation regression in the video-overview pipeline, possibly from a stronger framing bias or safety and face-handling change.

• –Google’s docs describe Video Overviews as AI-narrated slides built from notebook sources, with customizable format, language, and visual style, so composition quality is now a core part of the product experience.
• –The user report suggests the model is over-optimizing for “cinematic” framing at the expense of visible faces, which would be a real UX problem for explainers and educational content.
• –If this is a rollout change, the lack of obvious public discussion may simply mean most users are still testing it on varied sources, while creators notice the defect faster.
• –The most likely explanations are prompt drift, a model swap, or a stricter face-safety or framing policy that unintentionally degrades output.
• –If confirmed, this is the kind of regression that hurts trust quickly because it changes the visual semantics of the output, not just the polish.