Hand an AI clip to an editor: MP4, PNG frames and WAV on Sume

For a finishing editor, Sume can hand over three file types: an MP4 per clip, PNG stills at the source size, and a sample-exact WAV of the audio. Sume's docs do not describe HDR or EXR output, so a pro handoff that needs them is not something to promise from Sume.

Luma lists native HDR with 16-bit EXR export for Ray 3.2 (read 2026-10-05). Sume does not run that model, so this page covers what you can actually deliver from a Sume job.

The three deliverables

Pick the options an editor wants

For stills, ask for png, because the docs call it lossless inspection, and leave out max_edge so frames keep the source size. For audio, keep the default WAV, which is sample-exact. For picture, use precision: exact on trim if you need a frame-accurate cut, and keyframe only when a stream copy is acceptable, because the cut can start a GOP early and the result reports actual_start_seconds.

Set output.fps on purpose. If you leave it out, Timeline 1.0 follows the rate of the sources, and a mismatch warns with output_fps_resamples_sources.

A stills-and-stem pull

Two calls produce a handoff pack for one clip already in your workspace, and both need an Idempotency-Key. The stills call is below. The audio stem is POST /v1/audio-detach; see the Audio detach page for its format, range, channels and sample_rate fields.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: pack-frames-001" \
  -d '{"video_url":"https://media.sume.com/artifacts/artf_demo/clip.mp4",
       "at":[0,2.5,5],"format":"png"}'

What not to promise

The docs name MP4 for video output and WAV or MP3 for audio. Trim re-encodes with libx264 and yuv420p in exact mode. They do not describe a 10-bit, HDR, EXR or ProRes path, so for a grade-heavy or HDR finish the honest handoff is the MP4 plus the stills and stem, and a note that the editor starts from an SDR 8-bit 4:2:0 source.

Check each file with video inspect, which returns probe facts, before you send it on.

A handoff folder

A tidy pack for each shot has four files: the MP4 from the generation job, a PNG of the first and last usable frames, a WAV of the clip audio, and a text note with the job id and prompt. Name them by shot number so an editor can relink. The probe from video inspect (duration, resolution, frame rate and audio presence) belongs in the note.

Do not rename Sume URLs to local paths in a script without keeping the originals, because artifacts are durable media.sume.com files tied to your workspace.

Where the limits are

Frames cap at 24 per call and 300-second sources. Trim reads up to 1800 seconds and returns up to 900. Audio detach returns up to 900 seconds per call. For a long master, split the work in ranges and keep a manifest, so the editor can see which file maps to which part of the timeline.

Before you build

Before you build, read the linked Sume docs page for the exact request fields, limits and prices, because those pages are the source of truth and can change. Run one short, cheap test with your own material first, check the output in a player and in your editor, and only then scale to the full shot list. Keep every job id and file you approve, so a later change never forces you to regenerate work that was already signed off. Note that this post describes Sume's catalog and tools; Sume does not run Luma Ray 3.2, and nothing here claims HDR or EXR output.