Draft first: Omni 1.1 Flash 360p or Seedance 2.5 at 480p on Sume

Google describes Gemini Omni 1.1 Flash's 360p draft mode as "up to 60% faster and at a third of the cost" of standard 720p (Google, read 2026-10-05). That makes 360p a proof tier for Omni: check framing, motion and the prompt before you pay for 720p, 1080p or 4K. Seedance 2.5 has no 360p tier on Sume; its lowest listed resolution is 480p. The honest comparison is that each draft only proves things about its own model.

What a draft can and cannot tell you

Unless a model's catalog entry shows seed: true, treat a draft at one resolution and a final at another as separate generations from the same prompt, not a re-render of the same take. Read the seed field in GET /v1/videos/models for each id.

• Good for: composition, subject placement, camera move, whether the prompt was followed.
• Weak for: skin detail, small text, fabric texture, fine product edges.
• Not transferable: a 360p Omni draft says nothing about how Seedance 2.5 would render the same prompt. Different models, different takes.

Resolution ladders on Sume

Both ids are in the Video Router catalog. Omni's native audio is always on; Seedance's audio behavior is listed per model in the catalog fields, so read them before you plan a silent proof.

A proof request

Send the Omni proof at 360p, check it, then re-send the same prompt at the final resolution. Each is a separate async job with its own idempotency key.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: omni-proof-360-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "A ceramic mug on a wooden table, steam rising, slow push-in",
    "resolution": "360p",
    "duration": 6,
    "aspect_ratio": "16:9",
    "mode": "async"
  }'

A proof workflow that saves money

Treat the draft as a gate with a checklist. Write three yes or no questions before you generate: is the subject framed as intended, does the motion match the prompt, and is anything in frame that should not be. If a draft answers no to any of them, fix the prompt and re-draft; do not upgrade resolution to save a bad idea. Only when all three are yes do you send the final.

Keep the final's prompt byte for byte identical to the approved draft. Small edits between draft and final are the most common reason the final differs from what the client approved.

• Three yes/no questions per draft.
• Do not change the prompt between draft and final.
• Keep each job id so you can compare stills from the same timestamps.

Which draft to run first

Run the cheaper draft for the model you will use for the final. If the final is Seedance 2.5, draft at 480p on Seedance 2.5, not on Omni; the Omni 360p draft proves the prompt for Omni only. If you are undecided between the two, draft the same prompt on both at their lowest resolution and let that, plus the price in the catalog, make the decision before you commit to a final.

Keep drafts short, for example the minimum duration the catalog allows (3 seconds on Omni, 4 on Seedance 2.5), unless the motion you are testing needs longer.

Check pricing for each resolution in GET /v1/videos/models (pricing_skus) before a batch. See Google's Omni 1.1 Flash post for the draft-mode claim.