🔥 LongCat-Video Technical Report

💡 The paper introduces LongCat-Video, a 13.6 billion parameter video generation model based on the Diffusion Transformer framework. The model is designed to generate high-quality long videos efficiently, which is a crucial step towards creating world models. LongCat-Video has a unified architecture that can perform multiple tasks, including text-to-video, image-to-video, and video continuation, using a single model.

The model achieves efficient long video generation through a coarse-to-fine generation strategy and block sparse attention, allowing it to generate 720p, 30fps videos within minutes. The coarse-to-fine generation strategy works by gradually increasing the resolution and detail of the video, both in terms of time and space. Block sparse attention is a technique that reduces the computational cost of the model by only attending to certain parts of the input data.

The model was trained using a multi-reward reinforcement learning from human feedback approach, which enables it to achieve performance comparable to state-of-the-art models. The use of multi-reward reinforcement learning from human feedback allows the model to learn from human evaluators and improve its performance over time.

The results show that LongCat-Video excels in generating high-quality long videos, maintaining temporal coherence and quality even in videos that are several minutes long. The model's efficiency and performance make it a significant contribution to the field of video generation, and the fact that the code and model weights are publicly available will accelerate progress in this area. Overall, LongCat-Video is a foundational model that takes an important step towards creating world models, which are complex models that can simulate and generate realistic videos and other types of data.

📅 Published on Oct 25, 2025

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2510.22200
• PDF: https://arxiv.org/pdf/2510.22200

🤖 Models citing this paper:
• https://huggingface.co/meituan-longcat/LongCat-Video
• https://huggingface.co/Nishant2414/LongCat-Video
• https://huggingface.co/fjkane/LongCat-Video-bf16

🚀 Spaces citing this paper:
• https://huggingface.co/spaces/cpuai/LongCat-Video-Avatar
• https://huggingface.co/spaces/multimodalart/LongCat-Video
• https://huggingface.co/spaces/armaishere/meituan-longcat-LongCat-Video

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

#VideoGenerationModels #DiffusionTransformer #LongVideoSynthesis #TextToVideoSynthesis #ImageToVideoGeneration

🔥 OSP-Next: Efficient High-Quality Video Generation with Sparse Sequence Parallelism, HiF8 Quantization, and Reinforcement Learning

💡 The paper introduces OSP-Next, an efficient text-to-video generation model that addresses the high computational costs of existing models. The problem with current models, such as Diffusion Transformers, is that they achieve strong video generation quality but have quadratic costs due to full attention. To solve this, OSP-Next combines sparse attention, parallelism, quantization, and reinforcement learning.

The method used in OSP-Next is a hybrid full-sparse attention architecture, where the sparse component is implemented with Skiparse-2D Attention. This mechanism applies token-wise and group-wise sparse attention along spatial dimensions, leveraging locality while maintaining compatibility with FlashAttention kernels. The authors also propose Sparse Sequence Parallelism, which partitions subsequences across ranks and switches sparse patterns through a single All-to-All communication. This approach reduces communication volume by 75% compared to Ulysses Sequence Parallelism.

Additionally, OSP-Next incorporates HiF8 quantization to enable stable joint training with 8-bit quantization and sparse fine-tuning. The model also applies Mix-GRPO post-training to improve the performance of the sparse model. The authors evaluate OSP-Next on various settings, including 5-second 720P and 5-second 768P, and achieve significant speedups on NVIDIA H200 GPUs and Ascend 950PR hardware.

The results show that OSP-Next achieves a VBench total score of 83.73%, surpassing the Wan2.1 baseline. The model achieves up to 1.64 times single-GPU speedup and over 1.52 times eight-GPU speedup on NVIDIA H200 GPUs. Furthermore, with only a 0.4% drop in VBench total score, OSP-Next-HiF8 achieves 1.69 times and 2.27 times speedups under the two settings on a single Ascend 950PR, demonstrating the efficiency and performance of OSP-Next across hardware platforms. Overall, the paper contributes to the development of efficient text-to-video generation models with high-quality video synthesis and reduced computational costs.

📅 Published on May 27

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2605.28691
• PDF: https://arxiv.org/pdf/2605.28691

🤖 Models citing this paper:
• https://huggingface.co/yunyangge/OSP-Next

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

#VideoGenerationModels #SparseSequenceParallelism #HiF8Quantization #ReinforcementLearningForVideo #TextToVideoSynthesis

🔥 DreamX-World 1.0: A General-Purpose Interactive World Model

💡 DreamX-World 1.0 is a general-purpose interactive text-to-video model that generates long-horizon content with camera control and scene persistence. The problem addressed by this model is the need for a controllable and interactive world model that can generate high-quality video content. To solve this problem, the authors introduced several new methods, including a lightweight variant of projective positional encoding called E-PRoPE, which retains projective camera geometry while applying camera-aware attention to spatially reduced tokens.

The authors also converted a bidirectional video generator into a few-step autoregressive world model using causal forcing, DMD-style distillation, and long-rollout training. This training process exposes the model to its own generated history, reducing style and color drift that accumulates across autoregressive chunks. Additionally, the authors introduced Memory-Conditioned Scene Persistence, which retrieves earlier views through camera-geometry-based retrieval, and residual recycling, which makes the conditioning path less sensitive to imperfect memory latents.

The model also includes Event Instruction Tuning, which adds composable event control, and reinforcement learning alignment, which recovers camera control and visual quality after distillation. To improve efficiency, the authors used mixed-precision DiT execution, residual reuse, 75%-pruned VAE decoding, and asynchronous pipeline parallelism, allowing the model to reach up to 16 FPS on eight RTX 5090 GPUs.

The results show that DreamX-World 1.0 achieves a camera-control score of 73.75 and an overall score of 84.76, outperforming HY-WorldPlay 1.5 and LingBot-World in overall score. The model's ability to generate high-quality video content with camera control and scene persistence makes it a significant contribution to the field of interactive world models. Overall, DreamX-World 1.0 is a powerful tool for generating interactive and controllable video content, with potential applications in a variety of fields, including gaming, simulation, and education.

📅 Published on Jun 15

🔗 Links:
• GitHub: https://github.com/huggingface
• Project Page: https://huggingface.co/papers?q=projective%20positional%20encoding
• arXiv: https://arxiv.org/abs/2606.16993
• PDF: https://arxiv.org/pdf/2606.16993

🤖 Models citing this paper:
• https://huggingface.co/GD-ML/DreamX-World-5B

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

#TextToVideoSynthesis #InteractiveWorldModels #VideoContentGeneration #ScenePersistence #CameraControlMechanisms

🔥 DomainShuttle: Freeform Open Domain Subject-driven Text-to-video Generation

💡 The paper introduces DomainShuttle, a method for open domain subject-driven text-to-video generation that achieves high fidelity and flexibility across different scenarios. The problem with existing methods is that they focus on maximizing subject fidelity in in-domain scenarios, which limits their ability to adapt to cross-domain scenarios where subject-irrelevant properties need to vary according to the text prompt. DomainShuttle addresses this issue by introducing domain-aware modeling and a dual RoPE scheme. The method uses Domain-MoT to decouple videos and reference features, and domain-aware AdaLN for domain-specific modeling of reference images. It also uses the Video-Reference DualRoPE scheme to enable precise subject-level spatial modeling, and Cross-Pair Consistent Loss to extract intrinsic subject features unaffected by irrelevant features. The results of extensive experiments show that DomainShuttle achieves significant performance improvements over existing methods, demonstrating high subject fidelity and generative flexibility across diverse open domain application scenarios. This means that DomainShuttle can generate high-quality videos that retain the key features of the subject while also allowing for flexible editing and adaptation to different styles, semantic combinations, or domain attributes. Overall, the paper proposes a novel approach to text-to-video generation that can flexibly shuttle between different domains, making it a valuable contribution to the field of open domain subject-driven text-to-video generation.

📅 Published on Jun 24

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.26058
• PDF: https://arxiv.org/pdf/2606.26058
• Project Page: https://cn-makers.github.io/DomainShuttle/

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

#TextToVideoGeneration #OpenDomainGeneration #SubjectDrivenGeneration #DomainAwareModeling #TextToVideoSynthesis