Video API

Task-style video generation (sora-2 / seedance-2.0 / wan2.7 / veo3 / happyhorse / kling-v3 / grok-imagine series)

Models

sora-2

OpenAI Sora 2 / 2-pro. Default 720x1280, 4s; 1792x1024 / 1024x1792 trigger a higher pricing tier.

seedance-2.0

ByteDance Seedance 2.0: three public tiers (standard / fast / mini). The mode (text / image / reference) is inferred from the media inputs in the request.

wan2.7

Alibaba Wan 2.7. Single model id, 720P / 1080P, 2..15s. Supports text-to-video, image-to-video (first / first+last frame), and video-continuation.

Google Veo 3 / 3.1. Multiple model ids — fast / quality official tiers plus preview and lean beta series. 720p / 1080p / 4K, 4 / 6 / 8s. Supports text-to-video and image-to-video (first frame, or first + last frame interpolation).

happyhorse-1.0

HappyHorse 1.0. Adds a video-edit mode on top of T2V / I2V / R2V — drop in metadata.video_url to restyle an existing clip with optional reference images and audio control.

happyhorse-1.1

HappyHorse 1.1. Single model id, 720P / 1080P, 3..15s. Supports text-to-video, image-to-video (first frame), and reference-to-video (1..9 reference images).

kling-v3

Kling v3. Single model id, 720p / 1080p / 4k, 3..15s. Supports text-to-video, image-to-video (first frame or first+last frame), optional audio, multi-shot (model_params.multi_shot) and element customization.

grok-imagine

xAI Grok Imagine. Single model id, 480p / 720p, 6..30s. Supports text-to-video and image-to-video (1..7 reference images), with an optional generation style mode.