Tsinghua University (Deyuan Liu, Jungang Li, Kaiwen Zheng, Kai Jiang, Jintao Chen, Yaole Wang, Ming Lin, Tao Lin) · Streaming Video Distillation

Few-Step Streaming Video Generation via Condition Shifting

APEX extends the APEX method to few-step streaming video generation by using condition shifting to estimate the velocity field inside the generator, removing the separate fake-score network used in Distribution Matching Distillation. Compared with S2-Avatar on 180-second videos, it reports slightly higher instruction fulfilment (3.856 versus 3.831) and visual aesthetics (0.641898 versus 0.632606). Separate measurements show 4.4% lower peak GPU memory and 9.2% shorter mean iteration time than DMD. Motion quality favoured S2-Avatar in the human study, and no licence or code availability is stated on the page.

Open the source ↗

Similar Video & Animation releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.