Luma uni-1 multi-panel storyboards vs Sume n images

Luma says uni-1 generates multi-panel sequences, such as storyboards, with consistent style when the prompt describes the panels: one request, one image of several panels. On Sume, n requests up to 10 separate images per call, and per-model ceilings are lower.

Luma facts are from its Models page; Sume facts are from Image models, read 2026-10-01.

How does Luma make a storyboard?

The Models page lists multi-panel output under image model strengths: describe the panels in the prompt and the model keeps a consistent style across them. It adds no panel count, layout parameter or limit.

What does n do on Sume?

A Sume request with n: 4 returns four images, not one four-panel image.

How do I keep style across Sume panels?

Generate the first frame, then pass a public HTTPS URL of it in input_references for the next frames, so each call has a visual anchor. The docs describe input_references as reference images for image-to-image; whether style carries is something to test per model.

You can also describe the panels in one prompt and see how the model lays them out; the docs make no promise either way.

curl https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-image-2","prompt":"Frame 2: the detective opens the door","n":2,"input_references":[{"type":"image_url","image_url":{"url":"https://example.com/frame1.png"}}]}' 

What next, once I have the panels?

Turn the stills into motion with the video tools; storyboard to video covers that route, and ChatGPT Image storyboards covers a one-model workflow.