Does Wan VACE support fine-tuning, or LoRA integration? Or we can just train LoRA for Wan 2.1 and apply to VACE

[matabear-wyx]

[hanzhn]

The published VACE models are all adapters, leaving the trainable weights of base Wan2.1-T2V model unchanged. So I think existing LoRAs for Wan2.1-T2V can be directly used together with VACE adapters.