POST /api/v3/contents/generations/tasks with text — or an image for image-to-video (Volcano Ark-compatible).
Task accepted instantly. Per-second pre-charge only, never overpay.
Poll the task endpoint. Settled to actual duration — failed tasks auto-refund in full.
Seedance-first, but every frontier model is one endpoint away.
Text-to-video & image-to-video on Seedance 2.0 / 2.0 Fast. Per-second billing: pre-charge on submit, settle on completion, auto-refund on failure.
Token-accurate metering, balance pre-checks, idempotent settlement. Every cent traceable.
OpenAI, Claude, Gemini, Kimi, GLM, DeepSeek and more — OpenAI-compatible, switch models without changing code.
Weighted load balancing, automatic failover and circuit breaking across upstream channels.
Continuously adding the latest models with a unified calling experience