Luma Dream Machine capabilities

Luma exposes Dream Machine's image and video generation through a REST API so the same Ray models available in the app can be driven programmatically. Generation is asynchronous: a request returns an ID, and that ID is polled for status until the asset is ready, or a callback URL receives status updates as the job moves through states such as dreaming, completed, or failed.

Ray3 is positioned by Luma as delivering studio-grade HDR through native high dynamic range color generation, producing vivid high-dynamic-range video from text prompts, SDR images, or SDR video. This is the signature differentiator versus most consumer AI video tools, which output SDR only.

Dream Machine bills by a credit system across web tiers: Free (limited credits, 720p, watermarked), Lite ($9.99/mo, 3,200 credits), Plus ($29.99/mo, 10,000 credits, no watermark), Unlimited ($94.99/mo, 10,000 Fast-Mode credits plus unlimited Relaxed Mode), and Enterprise (custom, 20,000 credits). Credits are the single currency for generating both images and video, and the per-generation cost varies by model and resolution. Luma's plans grant commercial use on paid tiers.

Ray3 Modify edits and reimagines existing footage while keeping a real human performance at the center of the workflow. Users direct with their physical actions and transform form, environment, and identity while preserving the original motion, emotion, and timing. Common uses include wardrobe swaps, environment changes, relighting, character transformations, and virtual product placement.

Luma Dream Machine generates short cinematic video clips from a text prompt, from a still image, or from text plus an image as a guiding frame, using the Ray3 model family. Ray3 is described as a reasoning-driven model that can think in concepts and visuals, evaluate its own output, and iterate to deliver better generations in fewer tries. The Ray3.14 update added native 1080p generation, roughly 4x faster performance, and lower per-generation cost.