Text, image, video, speech and embedding models behind OpenAI- and Anthropic-compatible APIs, routed by task difficulty.
文生视频与图生视频，失败按规则退款。
model/video-10s