Firefly's Kling Omni references vs Sume input_references

Adobe Firefly now takes reference images for Kling 3.0 Omni in its Video Generation tab and Boards. On Sume the same idea is input_references on POST /v1/videos (or reference_image_urls, 1 to 9 images, on the legacy Video 1.0 URL), and each model lists the reference types it accepts.

What Adobe announced

The Adobe community post for September 2026 says reference images were added for Kling 3.0 Omni in the Video Generation tab and in Boards, and that prompt length was raised for image and video on web. The page is a product update, not an API reference, so this post compares only the concept.

How Sume handles references

input_references provides style or content reference images for reference-to-video. The model uses them as visual guidance, not exact frames. frame_images is different: it sets first or last frames for image-to-video, and it wins if both are sent.

Only models whose supported_input_references lists a type accept that type. Audio and video references are honored by the Seedance 2.x models, Wan 3.0, MiniMax H3 and MiniMax H3 Max.

A reference request

Check the model first with GET /v1/videos/models, then submit.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ref-video-001" \
  -d '{
    "model": "seedance-2",
    "prompt": "The character from the reference walks through a market at dusk",
    "input_references": [{
      "type": "image_url",
      "image_url": {"url": "https://example.com/character.png"}
    }],
    "resolution": "1080p"
  }'

Reference URLs must be public HTTPS images. Sume's docs list no Kling model for this route, so do not assume Kling-specific behavior. Pick the model from the live catalog.