Rogue Studio pitches uncensored Rogue 1.0 as 'HBO of AI'

A studio called Rogue is describing itself as "a playground for creative ethical mischief" and pitching its in-house video generator, Rogue 1.0, as the "HBO of AI," according to Wired.

The product is an uncensored text-to-video model with native audio. Wired's profile says it accepts up to 13 character references and outputs 1080p at 24 to 60 fps. The company frames itself as an alternative to the major labs' content rules — Rogue Studio calls its offering "a playground for creative ethical mischief."

The architecture is the part worth reading twice. Rogue routes safe-for-work jobs out to Sora, Kling, Luma, Google, MiniMax and Seeddance, and keeps the NSFW work on its own model. In other words the front end is a router across the big video labs, and the in-house model exists specifically to cover what those labs will not generate. Deepfakes are said to be a terms-of-service violation.

Two of the AI experts we track shared the Wired piece this week.

A lot is unresolved from the profile alone: whether the six named model providers have blessed Rogue's use of their APIs, how the 13-reference likeness limit is policed in practice, and how the company verifies who is depicted in the images users upload.