4K AI video generation API: resolutions by model, 360p to 4K

To generate 4K video with the Sume API, send POST /v1/videos with model: "gemini-omni-flash-1.1" and resolution: "4K". It is the one catalog model that lists 4K in supported_resolutions, next to 360p, 720p, and 1080p. minimax-h3 can also return 2K or 4K, as upscales of its native 768p output that are priced if requested.

Resolutions come from the Video generation docs and the catalog behind GET /v1/videos/models, read on 2026-09-26.

Which resolutions does each video model support?

Each model advertises the subset it accepts in supported_resolutions, and POST /v1/videos checks resolution against that list. The range runs from 360p on Gemini Omni Flash 1.1 to 4K.

• 4K: gemini-omni-flash-1.1, and minimax-h3 as a priced upscale.
• 1080p: Seedance 2.x, wan-3.0, kling-3, minimax-h3-max, and gemini-omni-flash-1.1.
• 720p at most: grok-imagine-video-1.5.

How do I request a 4K clip?

Send resolution with a value the model lists. This adapts a docs prompt to a 4K Gemini Omni Flash 1.1 clip:

• Gemini Omni Flash 1.1 clips run 3–10 seconds in 16:9 or 9:16, with native synced audio always on.
• In its edit mode, resolution is optional and defaults to 720p; see Edit a video with a prompt.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: t2v-4k-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "A time-lapse of a flower blooming",
    "resolution": "4K",
    "aspect_ratio": "16:9",
    "duration": 8
  }'

What does 768p mean on MiniMax H3?

minimax-h3 renders natively at 480p or 768p, and the docs call 768p first-class, not 720p; a 720p request to it is refused. Its 2K and 4K outputs are upscales of 768p that do not appear in supported_resolutions. minimax-h3-max offers 480p, 768p, and 1080p, where 1080p is a latent refinement from native 768p on the same model. More on both in MiniMax H3 video API.

What happens if I ask for a resolution the model does not list?

POST /v1/videos refuses the request with a 400 and the error code unsupported_capability, for example “kling-3 does not support resolution 4K.”, and the error details carry the values the model accepts. On that route the accepted values match supported_resolutions with two additions: 2K and 4K on minimax-h3, and lowercase 4k on gemini-omni-flash-1.1.

Does a higher resolution cost more?

It depends on how the model prices. Wan 3.0, both MiniMax H3 models, and Gemini Omni Flash 1.1 price per video second with one rate per resolution, named per-video-second-<resolution> in pricing_skus, and on each of them the rate rises with resolution. Seedance models price per 1,000 video tokens. kling-3 and grok-imagine-video-1.5 list a per-video-second rate with no resolution tier.

minimax-h3 lists only its native 480p and 768p rates in pricing_skus; its 2K and 4K upscale rates appear in the model's constraints on GET /v1/video-router/models/minimax-h3. Every rate is the provider's list price × 1.25, plus a 5.5% agent fee by default; see API pricing.