Skip to main content

AI Model Catalog

Browse the video models available in AiuniVid. Compare providers, output formats, starting credits, and the best model for each creation workflow.

Model specifications side by side

Every value below is read from the same configuration the generator validates against, so this table cannot drift from what the API actually accepts. Upstream model IDs are BytePlus public catalog names — you can verify each model's limits in their documentation.

Seedance model specifications
ModelBest forDurationResolutionAspect ratiosImage inputCredits / second
Seedance 2.5dreamina-seedance-2-5-260628Longest clips, omni reference, audio-only input4–30s480P / 720P / 1080P16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptiveYes16.6
Seedance 2.0dreamina-seedance-2-0-260128Highest resolution, up to 4K4–15s480P / 720P / 1080P / 4K16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptiveYes11.4
Seedance 2.0 Fastdreamina-seedance-2-0-fast-260128Faster turnaround at lower cost4–15s480P / 720P16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptiveYes7.4
Seedance 2.0 Minidreamina-seedance-2-0-mini-260615Cheapest per second, for drafts4–15s480P / 720P16:9, 9:16, 1:1, 4:3, 3:4, 21:9, adaptiveYes5.8

Credits per second are quoted at each model's lowest resolution so the four are directly comparable; the generator quotes the exact cost for your chosen settings before anything is charged. 1080P output on Seedance 2.5 is H.265/HEVC — a few browsers need a download to play it.

Deciding between the two standard tiers? Seedance 2.5 vs 2.0: which to use compares them job by job.

Reference input limits

Reference-to-video accepts images, video, and audio together. These are the per-model ceilings the API enforces — exceed one and the request is rejected before it reaches the model, so nothing is charged.

Reference input limits by model
ModelReference imagesReference videosReference audioReference video lengthAudio-only inputFirst / last frame
Seedance 2.53010102–30sYesUp to 2
Seedance 2.09332–15sNoUp to 2
Seedance 2.0 Fast9332–15sNoUp to 2
Seedance 2.0 Mini9332–15sNoUp to 2

Reference video is capped on total length as well as per clip: Seedance 2.5 accepts up to 30 seconds across all clips combined, the 2.0 family up to 15. Audio-only generation — no image, no video — works on Seedance 2.5 only.

What the models reject, and what to do instead

These are the rejections that actually happen in production, not a generic policy summary. Credits frozen for a rejected job are released automatically.

Photos of real people

Every Seedance model refuses to animate a recognizable real face, including your own and including images you hold the rights to. Switching models does not help — the check is upstream of all four. The compliant path is to create a character under Assets from that image, then reference the character instead.

Reference video under 2 seconds

A motion reference shorter than the minimum is rejected on submission. This is about the reference clip's own length, not the length of the video you are generating — a 5-second output from a 1.5-second reference still fails.

Deleted or still-processing assets

A reference to an asset that was deleted after you selected it fails upstream, and retrying will never succeed. Reopen the character picker and select again — stale selections are dropped automatically when it reloads.

Prompts that trip content review

Prompt-level rejections name the prompt, not your reference material. Rewording usually clears it; on Seedance 2.5 specifically, some prompts that fail there pass on Seedance 2.0.

AI Video Models · Seedance Cost Per Second | AiuniVid