Luma AI releases Ray3.2 with up to 16 keyframes per clip, 1080p 20-second video, HDR export, and a full API for studio production workflows. Here's what Hong Kong creators need to know.
Luma AI just dropped Ray3.2 — and this isn't your average point-release. Think 16 keyframes per clip, 20-second 1080p shots, HDR export, and a proper API that finally lets studios bake AI video directly into their pipelines.
Here's what changed and why it matters for Hong Kong agencies, filmmakers, and content teams.
What Makes Ray3.2 Different
The headline feature is frame-level control through keyframes — up to 16 of them per clip. Instead of writing a prompt and hoping for the best, you can now shape motion, pacing, and scene progression at a granular level. This is a big leap from the "type and pray" approach that still dominates most AI video tools.
Ray3.2 also supports 1080p output at up to 20 seconds per clip. Most generative video models tap out at around 5-10 seconds. Longer clips mean fewer cuts and more natural-feeling sequences — especially useful for narrative work like commercials, music videos, and social content.
Built for Professional Workflows
Luma designed Ray3.2 with production teams in mind. Key upgrades:
- 16 keyframes per clip — set multiple control points to direct motion and timing within a single shot - 1080p 20-second generation — longer continuous scenes without stitching - HDR and 16-bit EXR export — industry-standard formats for VFX, compositing, and colour grading - Enhanced performance tracking — preserves skeletal posture, gestures, and facial expressions across up to 8 faces frame-by-frame - Reframe tool — adjust aspect ratios, extend frames, or swap backgrounds while maintaining original lighting
These aren't nice-to-haves for professional work. They're table stakes. Any Hong Kong post-production house that's tried to use AI video in a real pipeline has run into the "good luck matching your previous frame" problem. Ray3.2's keyframe system directly addresses that.
The API Changes the Game
The full Ray3.2 control surface is available through an API. That means studios can integrate AI video generation into their existing software stack — asset management systems, editorial workflows, review tools — without switching to a separate interface.
For Hong Kong agencies working on tight client deadlines, this is huge. You can build AI video into your production pipeline as a native tool rather than a standalone step that requires manual handoffs.
What This Means for Hong Kong Creators
Hong Kong's production scene runs on efficiency. Agency teams juggle multiple clients, film crews work fast-turnaround commercial shoots, and content studios need consistent output at scale. Ray3.2's combination of directability, longer clips, and API integration makes it a genuinely useful tool for:
- Commercial production — storyboard-accurate pre-vis with keyframed motion - Social content — longer clips mean fewer cuts and better pacing - Post-production — HDR and EXR exports fit existing colour pipelines - Agency workflows — embed AI video directly via API instead of manual round-trips
Frequently Asked Questions
Q: Is Ray3.2 a brand new model or an update to Ray3? A: It's a major update to the existing Ray3 architecture. Luma added keyframe control, longer generation, HDR export, and API access — significant enough to be a 3.x release in most companies' numbering.
Q: Can I use Ray3.2 inside Cooly Studio? A: Cooly.ai continuously adds support for new AI video models. Keep an eye on Cooly Studio's model picker for Ray3.2 availability — it's exactly the kind of pro-focused model that slots into production workflows.
Q: How does Ray3.2 compare to Veo 3.1 or Kling 3.0? A: Veo 3.1 excels at cinematic quality, Kling 3.0 at character consistency, and Ray3.2 at directability. The 16-keyframe system gives Ray3.2 a unique advantage for scripted, storyboard-driven work where you need precise motion control.
Q: Is the API available now? A: Yes. The Ray3.2 API is live and production-ready. Developers can integrate it into proprietary studio tools, render pipelines, and third-party creative products.
Q: Do I need a powerful GPU to use Ray3.2? A: No — Ray3.2 runs on Luma's cloud infrastructure. You access it through their web interface or API, so any modern machine with an internet connection can generate professional-grade video.