domain 4dTL;DR: Our model can create a 4D avatar using a single open domain reference image. Our approach consists of two steps during the inference process. First, the DiT model generates aThen, by transforming back to the depth domain with updated velocity, we move the reflectors towards correct position while preserving zero