xAI's video generator now locks your character's face and voice

The biggest frustration with generative video has always been scene-to-scene consistency. You nail a scene, cut to another angle, and suddenly your character has an entirely new face and voice. xAI is tackling that issue head-on.

Locking down faces, voices, and products

xAI just upgraded Imagine Video 1.5 with image and voice references, native 1080p resolution, and standalone text-to-video generation.

Users can now pass in a character picture alongside an audio clip to preserve identical facial traits and vocal tone across clips. You can upload up to seven reference images per prompt, letting you lock down specific components like an actor, a branded product, or a setting while only changing the motion.

Image and voice references are rolling out first to SuperGrok Heavy and Plus members in the US on web and iOS, with broader tier access coming soon. Developers can tap into text-to-video, 1080p, and image references right now via the grok-imagine-video-1.5 API, with voice matching available on request.

Why it matters

If you produce narrative shorts or product commercials, this drastically reduces the need to re-roll generations until the face looks right. True creative workflows require persistence, not just one-off spectacle clips.

Allowing up to seven image references feels like a calculated move toward commercial video production. It lets creators maintain both an actor and specific physical inventory without relying on messy workarounds, though keeping voice references behind an API request suggests caution around audio compute and impersonation risks.