L
Jul 21, 2026LongCat-Video-Avatar 1.5 cuts inference to 8 steps — here's
LongCat-Video-Avatar 1.5 by Meituan: DMD2 8-step distillation, Whisper-Large-v3 lip sync, INT8 offloading. MIT-licensed. How to run AT2V loc
Jul 21, 20267 min read0 reactions0 comments