Ideogram API seed for repeatable results vs Sume's 400

On Ideogram, pass an integer seed and the docs say you get reproducible generation. On Sume's POST /v1/images, seed is in the schema but no model advertises it yet, so sending one returns 400 unsupported_parameter.

Ideogram facts are from its pose variants and Ideogram 4.0 pages; Sume facts are from Image models, read 2026-10-01.

How does seed work on Ideogram?

The pose variants page describes seed as an optional integer "for repeatable results". The Ideogram 4.0 page says "Set for reproducible generation", and its responses return the seed for each image, so you can read back the value that was used.

What does Sume do with seed?

Why is it a 400 and not ignored?

A request that sets a parameter the selected model does not list is rejected rather than silently dropped. Parameters a model accepts are published as capability descriptors, and the docs say output_compression and seed are in the schema but no model advertises them yet. A hard error means you never believe a run was seeded when it was not.

curl https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ideogram/ideogram-v3","prompt":"a red kite","seed":42}'
# expect HTTP 400 with code unsupported_parameter

How do I get a consistent look without a seed?

Do not count on identical pixels from a repeated request. Keep the stable parts of the request fixed (model id, prompt, aspect_ratio), and save the generated file when a result is a keeper. The same constraint applies to video; see Wan 3 seed and size for how that route reports it. Re-read the capability descriptors later, since the docs say no model advertises seed yet.