Sume Auto video has no 1:1 or 21:9: which model to pin instead

Auto in Sume's Agents Videos panel offers 720p or 1080p, 16:9 or 9:16, and 3, 4, 5, 6, 8 or 10 seconds; it has no square, 4:3, 21:9 or longer-than-10-second option. When you need one of those, pin a model: Wan 3.0 for square and long clips, MiniMax H3 for 21:9, Kling 3.0 for 4K or 15 seconds.

That is a limit of the Auto entry, not of Sume's catalog, and the video router docs say the same thing: Auto create controls default to 720p and 8 seconds, with 3 to 10 second clips at 16:9 or 9:16.

What can Auto do?

Auto is the default entry and its description is that Video Router chooses the serving model. It supports text-to-video, image-to-video with a start frame, an end frame, reference slots and the Edit mode. Audio is always on and there is no toggle, because the docs say the current Auto family always generates audio and the generate_audio field should be omitted.

Sume does not disclose which family served an Auto request, and the docs tell you not to infer it from the output. That is the trade: convenience in exchange for no pin.

What about the legacy Video 1.0 entry?

The panel still lists Sume Video 1.0 with a "Retiring soon" badge, and it shares Auto's controls. The Video 1.0 docs call it a compatibility alias: for new integrations use POST /v1/videos with model: "sume/auto". Receipts report sume/auto, and retired routing_preset values such as kling and grok are accepted and ignored.

So choosing Video 1.0 does not get you a Kling or Grok route. If you want Kling, pin Kling.

When should I use Auto anyway?

Use Auto when the shape is vertical or widescreen and the clip is short, and you care more about getting a result than about which engine made it. A 9:16 product teaser at 8 seconds is the default case. Use a pinned model when the shape, length, audio behavior or reference rules are part of the brief, or when you need the same model next week for a consistent look.

The API makes the pin explicit: model: "sume/auto" lets Sume pick, any catalog id pins. Resolution is a pure function of the normalized request and the catalog version, so an idempotent replay routes identically.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"sume/auto","prompt":"Vertical UGC product clip, natural light","aspect_ratio":"9:16","duration":5}'

What should I check before a batch?

Auto's lists are short and stable, but the pinned models differ from each other, so confirm the ratio and duration you need on the exact model id you plan to send.

Submit one short clip first, open the finished job, and compare what you asked for with what came back. Use an Idempotency-Key on each attempt, because a replay with the same key returns the original job instead of creating and billing a second one. Only then queue the rest.

Sume bills provider list times 1.25 on every video model, so budget from the finished job's usage and cost fields rather than from a panel estimate.

What does Sume not do?

Sume does not widen Auto's shapes on request, and it does not let you ask Auto for a specific family. The decision guide in Auto or pinned and the Auto default post cover the rest.