Generate video from audio and a reference image. This app uses a distilled model; for the full version, deploy the open-source model.
Inference Resolution, default: 480P(推理分辨率，默认480P)