Midjourney alternative for product stills by API: what to send on Sume

Sume does not list a Midjourney model in its image catalog, so a product still on Sume is made with another model: GPT Image 2.5 for edits with masks and many references, Nano Banana 2 or Pro for reference edits and wide ratios, FLUX.2 pro for general work, or Seedream 5.0 Lite for cheaper runs (Image API). We have not compared these to Midjourney's output; the plan below lets you do it on your own products.

Which id for which job

Product stills are an edit job more than a text-to-image job: you have a real packshot and want a new scene. That favours models that list input_references. Start with the table, then test two.

Run the test

Pick three of your products, one easy, one glossy, one with fine print. Send each through two models with the same prompt and a preserve list. Compare with a checklist: label text, edges, shadow, colour. Keep the first-pass cost in mind: the table's rates are list times 1.25, rounded up to a cent, and failed generations are not billed.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "openai/gpt-image-2.5",
  "prompt": "Place the product from Image 1 on a travertine ledge, soft window light. Keep the product exactly as photographed. No added text.",
  "input_references": [{"type": "image_url", "image_url": {"url": "https://example.com/packshot.jpg"}}],
  "aspect_ratio": "auto"
}''

If you rely on Midjourney's look

A look is a taste, not a spec. If a team needs that exact aesthetic, keep using it for concepts and use Sume's models for the packshot-based final. A related post explains what Sume does and does not offer around Midjourney (Midjourney API on Sume).