sume/auto and idempotent retries: same price, same route on replay
New video models land weekly. With model sume/auto and an Idempotency-Key, a retry of a Sume video submit gets the original job, price and route.

Send model: "sume/auto" with an Idempotency-Key, and a retry of the submit returns the same job at the same price on the same route. The Sume video docs state that resolution is a pure function of the normalized request and the catalog version, so a replay cannot drift to another family when the catalog changes between your first call and your retry.
Why catalogs churn
The Runway API changelog, read today, shows how fast a video catalog moves: Grok Imagine Video 1.5 Lite on 2026-10-01, Seedance 2.5 Draft Mode on 2026-09-28, and MiniMax H3 Max on 2026-09-03. If your code pins a model id, each addition is a decision. If it sends sume/auto, Sume makes it, and the poll response reports sume/auto as the model.
| Date | Runway change | Sume side |
|---|---|---|
| 2026-10-01 | Grok Imagine Video 1.5 Lite, 1 to 15 s | grok-imagine-video-1.5 is in the Sume list |
| 2026-09-28 | Seedance 2.5 Draft Mode, 480p preview | seedance-2.5 accepts 480p, 720p, 1080p |
| 2026-09-03 | MiniMax H3 Max, 5/8 credits per second | minimax-h3-max, 480p to 1080p |
What Sume promises
Sume never discloses which family served an auto request, so do not infer it from the output. A replay with the same key and payload returns the original job; the same key with a different payload is 409 idempotency_conflict. A job is readable only by the member whose key created it.
If you need a specific family, pin its id from GET /v1/videos/models, which lists durations, resolutions and generate_audio for each.
The request
Use a stable key per logical request, such as an order id, not a fresh uuid on each retry.
curl -X POST "https://api.sume.com/v1/videos" \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: order-8823-hero-v1" \
-d '{
"model": "sume/auto",
"prompt": "A vertical product clip on a desk, natural light",
"aspect_ratio": "9:16",
"duration": 5
}'Retry rules
Retry on 429 rate_limited using retry-after, on 429 queue_full after jobs finish, and on 503 provider_capacity_exceeded later. Keep the same key each time. Do not retry a 402 insufficient_credits until the balance changes.
Pin or auto
Pin a model when the look must match across a campaign, when you need a specific duration the catalog lists for only some models, or when audio and reference types matter: the Seedance 2.x family, Wan 3.0 and the MiniMax H3 models accept audio and video references, while Gemini Omni Flash 1.1 and H3 Max Recast accept video but not audio.
Use sume/auto when you care about the brief and the budget, not the family. Responses name sume/auto, so your logs record that choice and not a guess.
The Runway rows show that vendors add and rename models often, but they are Runway facts and do not mean that Sume lists the same ids. Sume does not list Veo 3.1 in the v1 video catalog. Read GET /v1/videos/models for what is live today and treat any other list as a lead.
Sources
Related posts
More in Developers
- model sume/auto on /v1/videos vs pinning Seedance 2.5: what differs
sume/auto lets Sume pick the video family and never says which. Its documented limits are 3 to 10 s, so a 15 or 30 second Seedance clip must be pinned.
- Cancel a Sume bulk run: no queue endpoint, so cancel each child
Sume has no public cancel for a bulk queue. Poll the queue, then POST /v1/format-runs/{run_id}/cancel for each running child. fal cancels one request with PUT.
- Resume a bulk run after a crash: replay the Idempotency-Key
If your client dies after POSTing a 100-item bulk run, replay the same key and same body: Sume returns 202 with the original queue. A new body gets 409.
- Sume Image API 400 unsupported_parameter: which field fails where
Which Sume image models list quality, resolution, mask_url, background, output_format and references, and which never do (seed, stream). Plus a check script.
Written by Sume