LongLive-Plug: Once-for-All Distillation for Video Generation
LongLive-Plug distills reusable LoRAs so 54 video models gain sampling and guidance capabilities.
LongLive-Plug is a once-for-all distillation framework that learns reusable video-generation capabilities as LoRAs on a base model for training-free transfer to compatible downstream models. The adapters cover single-pass classifier-free guidance, few-step sampling, and long-context error correction for autoregressive generation, and remain reusable when models add conditioning or output channels. The authors verify deployment on 54 downstream models across three backbone families and eight task categories, including world modeling, robotics, editing, and multimodal generation.
- Capabilities are distilled once as reusable LoRAs on a base model.
- Adapters cover CFG, few-step sampling, and long-context error correction.
- A CFG LoRA still controls guidance via its inference weight.
- Training-free deployment was checked on 54 downstream models.
- Models span three backbone families and eight task categories.
Full article175 words · extracted from huggingface.co · click to collapse
Video diffusion models are increasingly developed into specialized models for diverse downstream tasks, and this development often includes a distillation stage, for example to accelerate sampling or to improve long-video generation. This stage is typically repeated for every specialized model. We introduce LongLive-Plug, a once-for-all distillation framework that learns reusable capabilities as LoRAs on a base model for training-free, plug-and-play deployment to compatible downstream models. These capabilities include single-pass classifier-free guidance, few-step sampling, and long-context error correction for autoregressive generation. The adapters remain reusable even when downstream models add conditioning branches, expand output channels. Despite training at a fixed guidance scale, our dedicated CFG LoRA provides text guidance control through its inference weight. Combining it with a few-step LoRA simultaneously preserves few-step generation and CFG controllability on downstream tasks. We verify training-free deployment on 54 downstream models across three backbone families and eight task categories, including world modeling, robotics, editing, and multimodal generation. The approach may support additional compatible models. Each capability can thus be distilled once per backbone family and reused without per-target retraining.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.38154