Alibaba: wanx2.1-vace-plus API

Wanx2.1-vace-plus is the world's first unified open-source foundation model for video generation and comprehensive editing. Built on a 14-billion parameter architecture, it integrates text-to-video, image-to-video, video-to-video, and localized editing into a single DiT (Diffusion Transformer) framework. Unlike standard models, VACE supports multi-modal inputs (Text + Image + Video + Mask + Control signals) simultaneously, enabling complex workflows like "replace this actor with this specific character image while keeping the background and motion consistent." It represents a paradigm shift from "one model per task" to an all-in-one creative suite for Hollywood-level video manipulation.

• Input: text, image, video
• Output: video
• File input: Supported
• Released: 2025-05-14
• Knowledge cutoff: 2025-03-31