--- title: "Luma's Ray3.2 Lets Directors Pin 16 Keyframes to Control AI Video" subtitle: "Luma's Ray3.2 brings 16-keyframe control, 8-face tracking, 20-second 1080p clips, and a first-ever API to professional video pipelines" url: "https://alphasignal.ai/news/luma-s-ray3-2-lets-directors-pin-16-keyframes-to-control-ai-video" author: "AlphaSignal Newsroom" author_url: "https://alphasignal.ai/editorial-team" company: "Luma" company_url: "https://alphasignal.ai/company/lumalabsai" published: "2026-06-09T14:46:14.000Z" updated: "2026-06-16T20:12:38.096Z" type: "News" topics: - "Video" - "Benchmarks" subtopics: - "Video Generation" - "Video Understanding" upvotes: 2912 read_time: "6 min" source_url: "https://lumalabs.ai/ray3-2" comments: 0 --- # Luma's Ray3.2 Lets Directors Pin 16 Keyframes to Control AI Video > Luma's Ray3.2 brings 16-keyframe control, 8-face tracking, 20-second 1080p clips, and a first-ever API to professional video pipelines [Ray3.2](https://lumalabs.ai/ray3-2) is Luma's sharpest answer yet to a question the AI video field has been circling for two years: what does it actually mean to *direct* a generated video, rather than just prompt one? The new model, designed in collaboration with creatives from the entertainment, advertising, and gaming industries, moves the control surface from a text box to something closer to a real production timeline. ## From prompting to directing Ray3.2 allows users to direct AI video frame by frame. With this update, creators are no longer just prompting , they are directing. That framing isn't just marketing. The headline capability is **Multi-Keyframe**, which lets you pin up to 16 reference frames inside a single clip. Think of keyframes here the same way an animator does: anchor points that tell the model exactly what the scene should look like at specific moments, with the model filling in the motion between them. By enabling the placement of up to 16 keyframes within a single clip, Luma grants directors and production teams absolute authority over pacing and motion. Previously, you could set a start and end frame and hope the model interpolated something usable. Now you can choreograph a camera move, a character gesture, and a lighting change all within one generation pass. ## What's actually new - **Multi-Keyframe (up to 16):** Pin exact visual states at any point in the clip. Control camera paths, narrative beats, and visual progressions without re-generating. - **Expressive Facial Performance:** Creators can choreograph exact narrative beats, camera paths, and visual progressions. Enhanced Performance Tracking and Expressive Facial Performance ensure that the nuance of a scene is accurately captured and transferred, maintaining skeletal posture and gestures, while tracking the full expressive state for up to eight faces simultaneously, frame by frame. - **Modify Video V2:** Reshapes existing footage , swap backgrounds, change environments, restyle wardrobe , while preserving the original lighting and performance. Clips up to 20 seconds at 1080p. - **Reframe:** Change aspect ratios, extend the frame, or replace a background after the fact, without reshooting. One generation, every platform format. - **HDR + EXR Export:** HDR generation continues to be a strength, and paired with EXR outputs it's ready for professional-grade industry workflows. The EXR frames are in ACES2065-1 (AP0) color space , the standard used in Hollywood color pipelines. - **Motion Transfer and Camera Motion Transfer:** Apply the movement dynamics from a reference video to new subjects, scenes, or styles. Camera moves can be transferred across entirely different worlds. ## The API is the real unlock for builders For the first time, Ray3.2 is available as an API, allowing developers to integrate Luma's professional-grade video generation directly into enterprise products, proprietary tools, and custom workflows. This is a significant shift. Until now, the full control surface of the Ray model , keyframes, modify, reframe, HDR , was only accessible through Luma's own web UI. Now it can be wired into render farms, in-house production tools, or any custom pipeline. Luma's stated design goal for the [API](https://lumalabs.ai/api) is zero middleware: it's built to slot into existing tech stacks without glue code. For teams already running automated content pipelines, that matters a lot. The API uses a credit system rather than a subscription tier, with costs scaling by resolution and output format. ## What it costs Ray3.2 text-to-video and image-to-video generation at 1080p runs 400 credits for 5 seconds and 1200 credits for 10 seconds. HDR output doubles the credit cost, and HDR with EXR export triples it. The Luma Agents subscription bundles Ray3.14, Uni-1, Veo 3.1, Kling 3.0, Seedance, and ElevenLabs audio under one credit pool , Plus at $30/month, Pro at $90/month, Ultra at $300/month. The Luma API does not have a subscription tier , it uses a credit system. ## Where it fits in the competitive field The AI video generation market is crowded and moving fast. Runway Gen-4.5 leads technically, but specialized models like Veo 3 for audio, Luma Ray3 for HDR, and Kling for unified workflows each dominate their niches. Luma's bet with Ray3.2 is that the professional production segment , studios, agencies, post-production houses , cares more about control and pipeline compatibility than raw quality scores on leaderboards. Luma is positioning Ray explicitly as professional infrastructure priced for production volume rather than a consumer app. The company's $900M Series C led by HUMAIN, and enterprise Luma Agents deployments at Publicis, Adidas, and Mazda all reinforce this direction. Ray3.2 is the clearest expression of that strategy yet: every new feature in this release , EXR export, multi-keyframe, performance tracking , is something a post-production team would ask for, not a casual user. The one notable gap: Veo 3.1 is the strongest current default if synchronized audio matters. Most of the rest of the field is still primarily video-first, with audio added in a separate step. Ray3.2 does not generate native audio, which means any dialogue or sound design still requires a separate pass. ## Practical use cases - **Advertising and brand campaigns:** Generate market-specific variations of a hero clip , different backgrounds, seasons, product placements , without reshoots. The Reframe tool handles format delivery across platforms in one pass. - **Film and TV pre-production:** Animatics and previs from storyboards, using keyframes to match the exact shot composition a director has already approved. - **VFX and compositing pipelines:** 16-bit EXR outputs drop directly into Nuke or DaVinci Resolve without any conversion step. The ACES color space means no re-grading from scratch. - **Character performance transfer:** Transform an actor's performance into a different character, environment, or visual style while preserving the original facial expressions and body language. - **Developer tools:** Build video generation into internal creative tools, render farm automation, or SaaS products via the new API , without managing a separate UI layer. ## The bigger shift What Ray3.2 signals is a maturation in how the field thinks about controllability. Early generative video was defined by the quality of what the model could produce on its own. The next phase , where Ray3.2 is squarely positioned , is about how precisely a human can steer that output. By unlocking the ability to dictate story beats and guide the motion of a scene, Ray3.2 empowers professional creators to execute their exact vision, knowing the technology will follow their lead. The assumption that needed updating was that AI video generation and professional production pipelines were fundamentally incompatible , that you could get impressive demos or you could get pipeline-ready assets, but not both. For HDR color-managed pipelines, Luma Ray3 is now the go-to choice in the field. Ray3.2 pushes that further: it's available now in the [Dream Machine](https://app.lumalabs.ai) web app and via the [Luma API](https://lumalabs.ai/api). ## Sources - [X post by @lumalabsai](https://x.com/lumalabsai/status/2064358414143143954) - [Luma Introduces Ray3.2 Model & API: Complete Creative Control for Video Generation](https://lumalabs.ai/news/introducing-ray-3-2) - [Ray3.2 Official Product Page — Luma](https://lumalabs.ai/ray3-2) - [Luma AI Pricing 2026: Plans, Costs & Real Budget](https://checkthat.ai/brands/luma-ai/pricing) - [Best Video Generation AI Models in 2026 — Pinggy](https://pinggy.io/blog/best_video_generation_ai_models/) - [AI Video Stack 2026: Live Comparison & Rankings — RCTV](https://rctv.com/posts/ai-video-stack-2026/)