🔥 SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer

💡 The paper introduces SANA-Video, a small diffusion model designed for efficient video generation. The problem addressed is the high cost and slow speed of existing video generation models. To solve this, the authors propose two core designs: Linear DiT, which leverages linear attention as the core operation, and a constant-memory KV cache for block linear attention. This cache provides global context at a fixed memory cost, eliminating the need for a traditional KV cache and enabling efficient long video generation.

The method used is a block-wise autoregressive approach for long video generation, which employs a constant-memory state derived from the cumulative properties of linear attention. The authors also explore effective data filters and model training strategies, which narrow the training cost to 12 days on 64 H100 GPUs, a significant reduction compared to other models.

The results show that SANA-Video achieves competitive performance compared to modern state-of-the-art small diffusion models, while being 16 times faster in measured latency. The model can generate high-resolution, high-quality videos up to 720x1280 resolution and minute-length duration at a remarkably fast speed. Additionally, SANA-Video can be deployed on RTX 5090 GPUs, accelerating the inference speed of generating a 5-second 720p video from 71 seconds to 29 seconds, a 2.4 times speedup. Overall, SANA-Video enables low-cost, high-quality video generation, making it a significant contribution to the field of video generation.

📅 Published on Sep 29, 2025

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2509.24695
• PDF: https://arxiv.org/pdf/2509.24695
• Project Page: https://nvlabs.github.io/Sana/Video

🤖 Models citing this paper:
• https://huggingface.co/Efficient-Large-Model/SANA-Video_2B_720p
• https://huggingface.co/Efficient-Large-Model/SANA-Video_2B_480p
• https://huggingface.co/Efficient-Large-Model/SANA-Video_2B_480p_diffusers

🚀 Spaces citing this paper:
• https://huggingface.co/spaces/helenai/check-optimum-intel-support

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#VideoGenerationModels #DiffusionTransformer #BlockLinearAttention #EfficientVideoProcessing #AutoregressiveVideoGeneration

🔥 FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization

💡 The paper introduces FashionChameleon, a real-time and interactive framework for human-garment video customization in autoregressive video generation. The problem addressed is the inability of existing approaches to support low-latency and interactive garment control, which is crucial for applications such as e-commerce and content creation.

To solve this problem, the authors propose a method that consists of three key techniques. First, they train a Teacher Model with In-Context Learning on a single reference-garment pair, which encourages the model to implicitly preserve coherence during single-garment switching. Second, they introduce Streaming Distillation with In-Context Learning, which fine-tunes the model with in-context teacher forcing and improves extrapolation consistency via gradient-reweighted distribution matching distillation. Third, they propose Training-Free KV Cache Rescheduling, which includes garment KV refresh, historical KV withdraw, and reference KV disentangle to achieve garment switching while preserving motion coherence.

The results show that FashionChameleon uniquely supports interactive customization and consistent long-video extrapolation, while achieving real-time generation at 23.8 FPS on a single GPU. This is 30-180 times faster than existing baselines. The framework enables users to interactively switch garments during generation, making it a significant contribution to the field of human-centric video customization. Overall, the paper presents a novel approach to achieving real-time and interactive human-garment video customization, which has significant commercial value and potential applications in e-commerce and content creation.

📅 Published on May 15

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2605.15824
• PDF: https://arxiv.org/pdf/2605.15824
• Project Page: https://quanjiansong.github.io/projects/FashionChameleon/

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#RealTimeVideoCustomization #HumanGarmentInteraction #AutoregressiveVideoGeneration #InteractiveGarmentControl #EcommerceVideoTechnology

🔥 OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators

💡 The paper proposes a method called On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators, or OPSD-V, which aims to improve the quality of videos generated by few-step autoregressive video diffusion models. The problem with existing models is that they can produce long videos with low latency, but the quality of the video degrades over time due to error accumulation and weakened motion dynamics.

To address this issue, OPSD-V introduces real long-video data as temporal context during training, providing dense trajectory-level supervision to improve visual quality and motion dynamics. The method works by having a student model follow the exact inference-time rollout, generating each chunk of the video conditioned on its own previously generated cache. In parallel, a teacher model is evaluated at the same denoising states, but uses a cleaner temporal cache that can be replaced by real-video context. This provides corrective targets under on-policy cache dynamics, without changing the inference mechanism.

The results show that OPSD-V consistently improves the visual quality, motion dynamics, and VBenchLong scores of the generated videos. The method is applied to representative few-step autoregressive video models, including Self-Forcing and LongLive, and the experiments demonstrate significant improvements. A user study with 10 participants also shows that OPSD-V is preferred over the base models in 66 percent of overall-preference judgments, and 82.5 percent excluding ties. Overall, the paper contributes a novel method for improving the quality of videos generated by few-step autoregressive video diffusion models, without altering the inference mechanism.

📅 Published on Jul 9

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2607.08766
• PDF: https://arxiv.org/pdf/2607.08766
• Project Page: https://meigen-ai.github.io/OPSD-V/

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#AutoregressiveVideoGeneration #VideoDiffusionModels #SelfDistillationTechniques #FewStepVideoGeneration #PostTrainingOptimization