[Incipient]

Outputting video of that quality/consistency at 1 minute, for a 2.6B model seems insane?

[tredre3]

It's because it is insane/misleading. It's a two stage process, scroll to the key features:

> A dedicated 17B long-video refiner sharpens texture, motion, and late-window quality on top of the long-rollout backbone.

[basilgohar]

It's a very specific use case. This model can generate 1 minute videos of what is essentially a streaming game scene.