Midjourney --tile seamless pattern API: what Sume accepts

Sume's image API has no --tile equivalent. The request fields are listed per model, and a field a model does not list is rejected with 400 unsupported_parameter, so a tile field would fail rather than be ignored. You can still request pattern-style images with a prompt, but Sume's docs do not promise that the edges will repeat.

The Midjourney change is from its Sep 24, 2026 update post; the Sume rules are from Image generation and editing, read 2026-09-30.

What did Midjourney change?

The post says --tile now works better for V8.1 and V8.2: previously you might have seen seams between tiled images, and now they blend invisibly from one tile to the next. That is a Midjourney feature; the post does not describe an API.

What can a Sume image request set?

How do I ask for a pattern anyway?

Describe the pattern in the prompt, use aspect_ratio: "1:1", and request several images with n so you can pick candidates. Then check the edges yourself: place the output next to a copy of itself and look at the seam. Treat any result as a candidate, not a guaranteed tile.

Where do I read what a model accepts?

Call the model discovery route and read supported_parameters for the model you plan to use, as the docs advise, before pinning a ratio. For batches of candidates, see bulk image generation.