Video models

seedance2_5 generates up to 30 seconds with a large reference budget for images, video and audio. It is the model to use when a single clip needs to carry a narrative.

It supports 12 width:height ratios across 480p and 720p only, defaulting to 1280:720. Durations run 4–30 seconds. Input and reference videos must be at least 480p, and combined input plus reference video is capped at 30 seconds.

Video-to-video accepts mode: "reference" (the default) or mode: "extend".

At 480p, the ratios 864:496 and 496:864 deliver standard 854×480 and 480×854 output rather than the literal pixel counts.

Cost: 30 credits/second at 720p, 20 at 480p, plus half that per second of input and reference video. Reference images and audio are free. Minimum 80 credits per generation.

gen4.5 is Runway's flagship video model, generating from text alone or from a starting image.

Text-to-video supports landscape 1280:720 and portrait 720:1280 only. Image-to-video adds 1584:672, 1104:832, 832:1104, 672:1584 and square 960:960.

Prompt images must have an aspect ratio between 0.5 and 2.

Cost: 12 credits/second, so a five-second clip is 60 credits ($0.60).

gen4_turbo takes an image and animates it at 5 credits/second, the joint-lowest rate in the catalogue. Use it for volume workloads where per-clip fidelity matters less than throughput.

It supports landscape 1280:720, 1584:672, 1104:832, portrait 720:1280, 832:1104 and square 960:960. Prompt images must fall between aspect ratios 0.5 and 2.358.

aleph2 takes a video plus a text or image prompt and returns an edited video. It is the video-to-video workhorse rather than a generator.

It preserves the input video's resolution, up to 1080p. Input videos must be 2–30 seconds at 30 FPS or lower.

Cost: 28 credits/second, minimum 56 credits per generation.

act_two drives a character in an image or video using a reference performance. Character images, character videos and reference videos must all fall between aspect ratios 0.5 and 2.358.

At 5 credits/second it is as cheap as Gen-4 Turbo.

veo3.1 and veo3.1_fast generate from text or an image, with or without audio. Audio roughly doubles the cost: 40 credits/second with, 20 without, on veo3.1; 15 and 10 respectively on veo3.1_fast.

First and last keyframes must fall between aspect ratios 0.5 and 2.

hailuo3 accepts ratio values adaptive, 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16, at 768P or 2K (default 2K). Durations run 5–15 seconds.

adaptive requires at least one reference image or video. Keyframe mode (position: first/last) and unpositioned reference-image mode cannot be mixed, and reference audio is only valid alongside unpositioned reference images, never keyframes.

Reference video is billed at the same per-second rate as output, capped at 15 seconds. Reference images cost 2 credits each.

grok_imagine_1_5 text-to-video supports 1:1, 16:9, 9:16, 4:3, 3:4, 3:2 and 2:3 at 480p, 720p or 1080p (default 480p), for 1–15 seconds.

It accepts up to 7 image references, addressable in the prompt as [Image 1], [Image 2] and so on, plus 3 audio references of 3–15 seconds each. Audio references require at least one image reference, and any request carrying image references is capped at 720p.

Image-to-video instead uses a resolution parameter and follows the input image's aspect ratio, supporting only a single first frame via promptImage.

happyhorse_1_0 text-to-video supports ten ratios across 720p and 1080p: 1280:720, 1920:1080, 720:1280, 1080:1920, 960:960, 1440:1440, 1108:832, 1662:1248, 832:1108, 1662:1248.

Image-to-video uses resolution (720P or 1080P, default 720P) and follows the input image's ratio. Prompt images must be at least 300px on each side and fall between aspect ratios 0.55 and 1.8.

gemini_omni_flash text-to-video and image-to-video output landscape 1280:720 or portrait 720:1280, always at 720p.

Video-to-video has no ratio parameter — output is 720p with orientation matched to the input, and input videos can be at most 10 seconds.