Open-weights video wave: which ids Sume lists, and which it does not
Sume's v1 video catalog lists Seedance, Wan 3.0, MiniMax H3, Kling 3 and Omni 1.1. It has no LTX id and no weights or LoRA fields. How to check before you plan.
Sume's v1 video catalog does not list an LTX id. As of this read it lists seedance-2.5, seedance-2, seedance-2-fast, seedance-2-mini, kling-3, wan-3.0, grok-imagine-video-1.5, minimax-h3, minimax-h3-max, gemini-omni-flash-1.1 and h3-max-recast, plus higgsfield-genjutsu when its provider is configured. The API calls hosted models, so there is no field for your own weights or LoRAs.
If a new model is announced, do not assume it is callable on Sume. Check the catalog.
What is and is not callable
This table covers the names that dominate the current video news. It states only what the Sume catalog says. For anything about a vendor's own release, read the vendor's page.
| Name you may be searching | On Sume's v1 catalog? | Id |
|---|---|---|
| Seedance 2.5 | Yes | seedance-2.5 |
| Wan 3.0 | Yes | wan-3.0 |
| MiniMax H3 and H3 Max | Yes | minimax-h3, minimax-h3-max |
| Kling 3.0 | Yes (prompt and frame video) | kling-3 |
| Gemini Omni Flash 1.1 | Yes | gemini-omni-flash-1.1 |
| LTX models | No id in the v1 catalog | none |
| Veo 3.1 | Not in the v1 catalog | none |
What to do if your plan needs a missing model
Pick the nearest capability that is listed. If you wanted a fast, cheap clip with sound, compare wan-3.0 at 480p and Omni at 360p. If you wanted long single shots, look at seedance-2.5 and wan-3.0, both up to 30 seconds. If you wanted to fine-tune a style with a LoRA, Sume has no such field, so you need to describe the style in the prompt or with reference images.
Do not route around the catalog by guessing an id. The generate route rejects ids that are not in GET /v1/video-router/models, and model_not_found comes back as a 404.
Re-check on a schedule
The catalog changes. A small cron that fetches GET /v1/video-router/models and diffs the id list will tell you the day a new id appears, which is a better signal than a headline.
A planning rule
Treat every name in a headline as unavailable until the catalog lists it.
This post makes no claims about any vendor's release terms or weights. For those, the vendor's own page is the only citation that counts; this page only reports what Sume's API lists.
- Fetch the catalog and look for the id before you design around it.
- Write your pipeline against capabilities, not against a model name.
- Keep a fallback id in config so a missing model does not stop the job.
- For any vendor claim, open the vendor's own page and cite it there.
Sources
Related posts
More in Models
- Lyria 3.5 tempo and duration control: Flow Music versus the Sume API
Google says Lyria 3.5 gives easier tempo and duration control in Flow Music. On Sume the same model has no duration field, so you steer length in the prompt.
- Lyria 3.5 better vocals and pronunciation: a listening checklist
Google says Lyria 3.5 improves vocals, lyrics and pronunciation. Check a Sume Music 1.0 take yourself: listen, transcribe it with STT, and compare the words.
- MAI-Voice-2.1 vs Flash: what $7 per million saves on a 60-second short
Flash lists at $15 per million characters and MAI-Voice-2.1 at $22. On a 750-character short that is a fraction of a cent. Math for 1, 30 and 3,000 shorts.
- MCP server for video analysis: what can an agent actually read?
Sume's remote MCP lets an agent probe a clip, sample stills, pull exact frames and transcribe audio. Semantic scene tools are dev-only. Limits and costs inside.
Written by Sume