Model · Luma AI · Proprietary

Ray3.2

Text-to-videoImage-to-videoVideo editProprietary
Open the AIContentDrop generator →
Arena rating
No votes yet
Crowd votes (Design Arena board)
1,996
Arena status
Pre-launch
Facts checked
2026-10-06

Vendor-sourced facts

Published facts for Ray3.2 — vendor-sourced, checked 2026-10-06
FactValue
VendorLuma AI [1]
LicenceProprietary (API/app only) [1]
Licence detailAPI only (lumalabs.ai/api) plus the Dream Machine web app; no downloadable weights [1]
ModesText-to-video, Image-to-video, Video edit [1]
Release2026-06-09 [1]
Max resolution1080p [1]
Max clip duration20 s [1]
Native audioNot stated [1]
AccessAPI, web app [1]
Vendor pagehttps://lumalabs.ai/ray [1]

Arena rating: no votes yet. Nothing on this page is a score, a rank or a measurement by us — every value is what Luma AI publishes, with the page it came from listed under Sources.

Crowd votes · external board

On Design Arena's video board Ray3.2 appears as “Ray 3.2”: Elo 1096 · #33 of 48 · win rate 42.0% ± 2.2% · 1,996 votes (838W 1,158L) · avg 52.2s.[2] These are their voters' numbers, not this arena's — our own arena has no votes yet.

Source: Design Arena (designarena.ai/leaderboard/video), checked 2026-10-21.

About Ray3.2

Ray3.2 is Luma AI's June 2026 model, reached through the Luma API and the Dream Machine web app. [1] Luma's pitch is control rather than raw scale: the announcement describes directing a generation frame by frame, which is a different proposition from a single prompt and a roll of the dice.

The stated envelope is 1080p and up to twenty seconds, covering all three modes. [1] Luma does not state native audio support on the pages we read, so that cell is Not stated rather than No — the distinction matters, because an unanswered question and a documented absence are different facts and only one of them is a limitation.

“Engineered for precision and absolute creative control, Ray3.2 allows users to direct AI video frame by frame.”

Luma AI on Ray3.2 [1]

Want hands-on context before trusting anyone's ranking? Generate your own clips with a multi-model workflow on AIContentDrop: connect the platform over MCP · look at the entry plan first.

Sources

  1. Luma AI — Ray3.2, vendor documentation https://lumalabs.ai/news/introducing-ray-3-2 — checked 2026-10-06
  2. Design Arena — video leaderboard https://www.designarena.ai/leaderboard/video — checked 2026-10-21

Each [n] above is the vendor page the fact beside the marker was read from, on the date shown. Where a vendor states nothing, the table says “Not stated” and no value is inferred from anywhere else.