Introduction
Seedance is ByteDance Seed’s multimodal video generation family. It supports text-to-video, image-to-video (first / first-and-last frame), multimodal reference generation, and face-consistent generation backed by an asset library. Unified access through the GravitexAI gateway:- Base URL:
https://api.gravitex.ai - Authentication:
Authorization: Bearer sk-xxxxxxxxxx - Content-Type:
application/json
Two API surfaces
Seedance is exposed through two parallel API surfaces on this platform. They offer the same capabilities and differ only in protocol style:Asset library data is shared between the two surfaces (the same asset groups and assets can be referenced from either), but video task routing is independent: query a task on the same surface where you created it.
Models covered by these docs
Latest models and pricing
Models iterate quickly — the console model gallery is the source of truth for the current list, pricing, and capability comparison.
Supported endpoints
Platform API
Video generation
Asset library
Official mirror API
Video generation
Asset library (all via
/api/v3/seedance?Action=...&Version=2024-01-01)
Asset library types
Typical flow
- (Optional) Create an asset group → create assets → poll until
status: active(usually 1–3 minutes) - Create a video generation task, referencing assets with
asset://<ASSET_ID> - Poll the task about every 5 seconds until
completed(usually 30–120 seconds) - If it
failedon moderation, call Get moderation result on the platform API
Reference
Platform API
/v1/video/generations and /v1/assets, OpenAI styleOfficial mirror API
Identical request / response to BytePlus Ark
Model gallery
Latest model list and pricing
Console
Create API keys and track usage