MODEL J LABS / RESEARCH NEWS / 24 SEPTEMBER 2026

Coherence needs a memory of the world.

Long-form generation raises a familiar creative-tool problem: keeping one world consistent across many decisions.

ORIGINAL RESEARCH / GOOGLE RESEARCH

Automating coherent long-form video generation

Published 24 September 2026 · Selected for 24–30 September 2026

The finding

Google Research presents a planning and orchestration framework for multi-shot video, built around Gemini and Veo. It addresses visual continuity and the cascading errors that can emerge when generated shots are treated independently.

Our take

For Aether, the interesting idea is persistent world state: identities, relationships, and scene rules that remain inspectable across creative actions. Our interpretation is that generation should participate in a shared scene model instead of producing disconnected assets.

The question we would test

Can a creator edit one relationship and preserve the rest of the scene across later generations? This is a direction for investigation, not a feature already available in our editor prototype.

Explore the Aether projectAll analysis