Video-Rate Streaming Stylization on a Vision-Aware MLLM-Conditioned Edit Diffusion: Asymmetric Batched Inference on a Distilled UNet + MLLM Text Encoder 文章

ArXiv CS.CV2026-06-05NEWSen作者: Yoshiyuki Ootani

查看原文 →

Video-Rate Streaming Stylization on a Vision-Aware MLLM-Conditioned Edit Diffusion: Asymmetric Batched Inference on a Distilled UNet + MLLM Text Encoder · 相关技术

相关技术

periodic conditioning-refresh schedule static-prompt caching batched text-encoder amortisation asymmetric side-stream / main-stream CUDA pipelining distilled edit U-Net Multimodal Large Language Models (MLLMs)ControlNet-LLLite diffusion U-Net