---
title: "Luma Opens Ray3.2 API so Developers Build Hollywood-Grade Video Into Any Product"
subtitle: "Luma's Ray3.2 API brings 16-keyframe control, 20-second V2V, and 16-bit EXR export to production pipelines at pay-as-you-go pricing"
url: "https://alphasignal.ai/news/luma-opens-ray3-2-api-so-developers-build-hollywood-grade-video-into-any-product"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "Luma"
company_url: "https://alphasignal.ai/company/lumalabsai"
published: "2026-06-09T16:50:05.000Z"
updated: "2026-06-10T20:28:39.952Z"
type: "News"
topics:
  - "Video"
  - "Api"
subtopics:
  - "Video Generation"
  - "Video Editing"
upvotes: 157
read_time: "7 min"
source_url: "https://lumalabs.ai/api"
comments: 0
---

# Luma Opens Ray3.2 API so Developers Build Hollywood-Grade Video Into Any Product

> Luma's Ray3.2 API brings 16-keyframe control, 20-second V2V, and 16-bit EXR export to production pipelines at pay-as-you-go pricing

Luma has announced [Ray3.2](https://lumalabs.ai/api), a significant update to its Ray3 video generation model designed in collaboration with creatives from the entertainment, advertising, and gaming industries. The headline news for builders: for the first time, Ray3.2 is available as an API, letting developers integrate Luma's professional-grade video generation directly into enterprise products, proprietary tools, and custom workflows. This is not a minor capability bump. It's Luma drawing a clear line between AI video as a toy and AI video as a production tool.

## From prompting to directing

With this update, creators are no longer just prompting , they are directing. Ray3.2 introduces frame-level control, giving production teams precise control over how action moves and changes from start to finish. The key mechanism enabling this is Multi-Keyframe. Think of keyframes as anchor points you place on a timeline: you define what the scene looks like at specific moments, and the model fills 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. Creators can choreograph exact narrative beats, camera paths, and visual progressions to flawlessly match meticulous storyboards and client expectations. Previously, most video models gave you a start frame and an end frame at best. Sixteen intermediate control points is a different category of control entirely.

## What's actually new in Ray3.2

- **Multi-Keyframe:** Up to 16 keyframes inside a single clip, directing what changes and what holds, beat by beat , and this sequence-level control is exposed at the API.
- **Video-to-Video (V2V) at 20 seconds:** Modify Video V2 reshapes footage you already have , swap the wall, the world, the wardrobe , while the lighting holds and the performance survives, up to 20 seconds at 1080p.
- **Native HDR and 16-bit EXR export:** Native HDR Generation and 16-bit EXR Export so AI work composites alongside live-action plates in users' Resolve or Nuke. EXR is the file format used by professional colorists and compositors , the fact that generated footage can now be exported in this format is what makes it a real production asset.
- **Expressive Facial Performance:** Expressive Facial Performance preserves the actor's read, tracking the full expressive state for up to eight faces simultaneously, frame by frame.
- **Reframe:** The enhanced Reframe capability allows teams to reshape a shot after the take. Post-production artists can instantly adapt aspect ratios for different platforms, extend the frame, or replace a background while perfectly preserving the original lighting. Client feedback becomes a targeted, efficient edit rather than a costly reshoot.

## Why the EXR export matters more than it sounds

In professional film and advertising, the bottleneck has never been about whether AI can generate "something." It's whether it can generate something usable. Studio pipelines demand imagery with wide dynamic range, deep color fidelity, and temporal coherence. Without 10-, 12-, or 16-bit precision, grading, compositing, and VFX integration simply break down. Banding, crushed shadows, or clipped highlights can render entire sequences useless, no matter how fast the AI can produce them.

Ray3's leap to native HDR EXR means generated plates and sequences can finally sit side-by-side with live-action footage technically. The output is 16-bit EXR frames for color grading, compositing, and VFX, in ACES2065-1 (AP0) , the color science standard used in major Hollywood productions. This is the technical bar that separates a demo from a deliverable.

## The API, in practice

Luma designed the API to fit into existing tech stacks seamlessly. Whether building for render farms, in-house tools, or external creative applications, there is no middleware to glue together. It's built for pipelines, not just previews. The API follows a standard async request-and-poll pattern. Here's a minimal example for a T2V generation at 1080p:

```
import requests, time
API_KEY = "your-luma-api-key"
BASE = "https://api.lumalabs.ai/dream-machine/v1"
headers = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}
# Submit generation
resp = requests.post(f"{BASE}/generations", headers=headers, json={
    "prompt": "A lone astronaut walks across a red desert at golden hour",
    "model": "ray-3-2",
    "resolution": "1080p",
    "duration": 10
})
gen_id = resp.json()["id"]
# Poll for result
while True:
    status = requests.get(f"{BASE}/generations/{gen_id}", headers=headers).json()
    if status["state"] == "completed":
        print(status["assets"]["video"])
        break
    time.sleep(5)
```

For multi-keyframe control, you pass an array of keyframe objects with timestamps and reference images into the request body. The [official docs](https://docs.agents.lumalabs.ai) cover the full schema. You can also pair Ray3.2 with Luma's Uni-1 image model: use Uni-1 for image generation and Ray3.2 for video generation to create richer multimodal creative workflows with greater consistency and control.

## Pricing

The API uses a pay-as-you-go model with no minimum commitment on the Build tier, and a dedicated Scale tier for production workloads that need guaranteed throughput and latency SLAs. Here's the pricing breakdown for the Build tier:

| Task | 540p | 720p | 1080p |
| --- | --- | --- | --- |
| T2V / I2V (5s) | $0.15 | $0.30 | $1.20 |
| T2V / I2V (10s) | $0.45 | $0.90 | $3.60 |
| V2V (5s) | $0.72 | $1.44 | $2.16 |
| V2V (10s) | $1.08 | $2.16 | $4.32 |
| Reframe (per second) | $0.06 | $0.12 | $0.36 |

A few important cost multipliers to keep in mind: HDR output is 2x the SDR price, and HDR with EXR export is 3x. API access is billed separately from consumer subscriptions , credits don't transfer between the Dream Machine app and the API, so you need to fund each wallet independently.

## What it's good at , and where it falls short

Ray3.2 delivers strong cinematic quality , filmic lighting, steady camera work, believable motion , and excels in Modify V2V transformations. The EXR pipeline is genuinely differentiated for post-production work. On the other hand, the model can struggle with very complex multi-subject action scenes compared to Sora. It's also worth noting that no version of Ray3 produces native audio , music, dialogue, and effects have to be added with external tools or in the editing stage. If built-in audio is a requirement, Google's Veo 3.1 or Kling 2.6 generate the soundtrack together with the video.

## The industry shift this signals

Ray3.2 was designed in collaboration with creatives from the entertainment, advertising, and gaming industries. That's not marketing language , it's a product strategy. Luma's models are used by leading entertainment studios, advertising agencies, and technology partners worldwide, including Adobe and AWS. The company is backed by HUMAIN, Andreessen Horowitz, Amazon, AMD Ventures, and NVIDIA.

The broader context matters here. The right moment to talk about Luma's positioning arrived with the shutdown of Sora. OpenAI had a billion-dollar partnership with Disney, and Disney itself reportedly learned of the discontinuation less than an hour before the public announcement. The message for anyone working with generative video is clear: provider stability matters as much as model quality. Luma is making a deliberate bet that the teams who will win are those who treat AI video as infrastructure, not a feature , and the Ray3.2 API is the clearest expression of that bet yet.

The practical use cases this unlocks are concrete: brand teams generating product variants across every market format from a single shoot, VFX studios using AI-generated plates that actually survive a colorist's session, and ad agencies iterating on scene edits without scheduling a reshoot. Unlike earlier AI video tools that focused on short clips or social content, Ray3.2 is engineered for cinematic depth, technical fidelity, and professional integration , reducing creative bottlenecks, accelerating iteration, and opening filmmaking to creators who want Hollywood-level storytelling without Hollywood-level budgets.

## Sources

- [X post by @lumalabsai](https://x.com/lumalabsai/status/2064389582997897216)
- [Luma Ray3.2 API — Official Page](https://lumalabs.ai/api)
- [Luma Introduces Ray3.2 Model & API — Official Announcement](https://lumalabs.ai/news/introducing-ray-3-2)
- [Ray3 brings 16-bit HDR generative video — fxguide](https://www.fxguide.com/quicktakes/ray3-brings-16-bit-hdr-generative-video-now-in-adobe-firefly/)
- [Luma Dream Machine: Ray3 and Ray3.14 Review (2026)](https://pasqualepillitteri.it/en/news/4026/luma-dream-machine-ray3-review-2026)
- [Luma Video Models Field Guide](https://lumalabs.ai/learning-center/articles/luma-video-models-field-guide)