Luma Ray3.2 20 s 1080p vs MiniMax H3 Max 15 s 1080p on Sume

Luma's Ray3.2 page lists 1080p clips of up to 20 seconds with up to 16 keyframes, but Sume does not run Ray3.2 and its video router catalog does not list it. The closest 1080p MiniMax option on Sume is minimax-h3-max, which makes 5 to 15 seconds at $0.20 per second.

So a 20-second 1080p request on Sume becomes two H3 Max clips for $4.00, or one wan-3.0 clip at 1080p for $5.00, and neither is a drop-in match for Ray3.2's keyframe control.

What Luma says

The Luma announcement (read 2026-10-05) lists Ray3.2 with up to 16 keyframes per clip, 1080p output, clips up to 20 seconds, native HDR with 16-bit EXR export, facial performance tracking for up to 8 faces, a Reframe tool, and a full control surface through the API. This post uses those as vendor claims only and does not restate Luma's pricing.

None of the Luma features are available through Sume. The Video Router catalog lists seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast, kling-3, wan-3.0, grok-imagine-video-1.5, minimax-h3, minimax-h3-max, gemini-omni-flash-1.1, h3-max-recast, and higgsfield-genjutsu when its provider is configured.

What Sume offers near that spec

For 1080p, the catalog lists seedance-2.5 (4 to 30 s), seedance-2 (4 to 15 s), kling-3 (4 to 15 s), wan-3.0 (2 to 30 s), minimax-h3-max (5 to 15 s) and gemini-omni-flash-1.1 (3 to 10 s, also 4K). Frame control is limited to first and last frames on rows that list start/end-frame image-to-video, such as minimax-h3-max; the notes do not describe multi-keyframe control.

How the Sume prices were computed

H3 Max is $0.16 list per second at 1080p, times 1.25, which is $0.20, so 20 seconds is $4.00. Wan 3.0 is $0.20 list at 1080p, times 1.25 is $0.25, so 20 seconds is $5.00. Gemini Omni Flash 1.1 is $0.15 list at 1080p, times 1.25 is $0.1875, so a 10-second clip is $1.88 after rounding up and two of them are $3.76.

What you give up

If your Ray3.2 plan relies on HDR or EXR delivery, on tracking up to eight faces, or on a dense keyframe timeline, this post found no matching feature in the Sume catalog notes it read. If your plan only needs a 1080p clip with sound and a pinned first and last frame, H3 Max covers it in 15-second pieces, with native stereo audio on each.

For a 20-second shot, note that two H3 Max clips will meet at a cut. Plan the cut on an action beat, and use the final frame of the first clip as the first frame of the second to keep the subject in place.

• Need one continuous 20 s take at 1080p: Wan 3.0 is the Sume row that reaches it.
• Need 15 s with stereo audio at $3.00: H3 Max.
• Need HDR/EXR: not something this catalog page lists.

How to choose for a 20-second plan

Decide first whether the 20 seconds must be one take. If it can be two beats, two H3 Max clips of 10 seconds each give you native stereo audio, first and last frame control and reference inputs for $4.00 at 1080p, and you can test the first beat at 768p for $1.00 before you pay for the second. If it must be one take, Wan 3.0's 1080p row is the Sume option that accepts 20 seconds in one job, at $5.00.

Whichever you choose, read the catalog row first. The Video Router docs say limits differ for each model and tell callers to read capabilities from GET /v1/video-router/models instead of assuming one envelope.