AI video HDR and EXR export: what Sume's outputs are
Luma says Ray3.2 has native HDR with 16-bit EXR export. Sume's docs describe MP4 outputs and no HDR or EXR; here is where that matters in a finishing pipeline.

Luma's Ray3.2 announcement, dated June 9, 2026, lists native HDR with 16-bit EXR export. Sume's docs describe no HDR or EXR output: its media tools return MP4 files, and exact trim re-encodes with libx264 and yuv420p. If your grade or composite needs EXR frames, do that step outside Sume.
What does Luma say about HDR and EXR?
Luma's announcement, read 2026-09-29, names native HDR and 16-bit EXR export as Ray3.2 features. The text saved for this page gives no color space, container detail or price for them, so none are stated here.
Does Sume export EXR or HDR video?
Not in the docs read. The catalog fields a video model advertises are its id, resolutions, aspect ratios, sizes, durations, frame images, input references, audio flag, seed, pricing and passthrough parameters. None describes bit depth, color space or an image-sequence export.
What do Sume's own tools hand back?
Each media tool in the docs returns an MP4, so a clip that passes through one is a delivery file, not a master.
| Tool | Output | Detail in the docs |
|---|---|---|
| Timeline 1.0 | One MP4 | Compiled to ffmpeg on the worker media runtime |
| Video trim | A new MP4 | exact re-encodes with libx264 and yuv420p; keyframe is a stream copy |
| Video filter | A new MP4 | Output inherits source geometry, frame rate and audio unless the program changes them |
How do I protect quality if I grade elsewhere?
Download the generated file before any Sume post step and keep it as the original. Use precision: "keyframe" on a trim only when a stream copy is acceptable, because a copy skips the re-encode; the cut may start a GOP early, so re-base against actual_start_seconds. Do color work and any EXR conversion in your own tools, on the original, and only then make the delivery MP4 you plan to publish.
See the trim docs for the cut fields and the video filter docs for what a pixel pass can change.
What is a safe handoff to a grading tool?
Keep the original as the master and treat anything Sume post-processes as a delivery copy. In practice that comes down to four habits.
- Download each generated file from
GET /v1/videos/{jobId}/contentand store it untouched. - Note the frame rate of each clip. A Timeline render that omits
output.fpsfollows the sources; a different rate repeats or drops frames. - Run
POST /v1/video-filter/checkbefore any pixel pass; it is free and creates no job. - Treat Timeline output as the delivery version, and rebuild it from the originals if the grade changes.
Sources
Related posts
More in Media tools
- How to flip a video horizontally with an API
To mirror a video left to right, run the hflip filter. Sume's video filter takes it as a one-line filtergraph and returns a new MP4.
- How to make a video montage: clips cut to one music track
To make a video montage, pick the music first, mark where the cuts go, then lay short clips over the track at those times in one render.
- Can AI make a song from a voice memo or a video?
Suno v6 says it can start from a voice memo, image or video. Sume's music API takes a text prompt and one optional image, so audio and video stay outside it.
- Mastodon video size limit: 99 MB, one video per post
Mastodon's docs allow a video of up to 99 MB (MP4, M4V, MOV or WebM), one per post, transcoded to H.264 at up to 1300 kbps. What that means for your export.
Written by Sume