Midjourney style reference API: input_references on Sume

Sume has no sref parameter. The way to carry a style is the documented image-to-image shape: send a prompt plus one or more public HTTPS reference images in input_references, on a model whose descriptor allows references.

Midjourney's side is from its changelog posts, read 2026-10-01; Sume's is from the image docs. This is a different mechanism, not a copy of srefs.

What do the Midjourney changelogs say about srefs?

The 9-16-26 and 9-23-26 posts describe work inside the Midjourney app: a Styles sidebar with live previews, a search drawer that shows Frequent and Recent srefs, and default parameters as a next step. They describe app features, and I found no API described on those pages.

How do I pass a style reference on Sume?

The input_references field is an optional array described as "Reference images for image-to-image". The docs' own example prompt is "make this scene look like a watercolor painting" with one image_url reference.

What are the limits on references?

Reference URLs must be public HTTPS; localhost, private-network and non-HTTPS URLs are rejected before submission. References are per-model: models whose input_references descriptor is {"min": 0, "max": 0} are text-to-image only and reject them. Read the descriptor from GET /v1/images/models first.

What should I do next?

Pick one reference that shows the look you want, write the style in the prompt, and run it on a model that lists references. For the broader question see does Sume offer a Midjourney API, and for the image-to-image basics image to image AI.