Ray 3.2

Ray 3.2 is a video generation model built around four core workflows: text-to-video, image-to-video, video editing (transforming existing footage while preserving motion and performance), and video reframing (changing aspect ratios without cropping).

Unlike older AI video tools, Ray 3.2 handles complex motion, lighting changes, and multi-subject scenes with consistency. You get native 1080p HDR output—16-bit color generated by the model, not post-processed—and can export as 16-bit EXR in ACES2065-1 color space, ready for Nuke or DaVinci Resolve.

The model supports up to 16 keyframes per clip, letting you direct exactly where motion shifts, lighting changes, or a beat resolves. It tracks up to 8 faces simultaneously, handles character transformation, environment swaps, relighting, product changes, and campaign localization. Video edits preserve the original actor's performance while you change environments, outfits, or lighting. Reframing fills newly exposed canvas areas with scene-consistent content that matches mood and lighting.

Built for creators, marketers, VFX studios, agencies, and filmmakers. Generate from your browser with no GPU required, or integrate via the Luma API. Credits are one-time purchases (no subscription) and never expire.