Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2512.13313 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912765160980480 |
|---|---|
| author | Kling Team Chen, Jialu Ding, Yikang Fang, Zhixue Gai, Kun Gao, Yuan He, Kang Hua, Jingyun Jiang, Boyuan Lao, Mingming Li, Xiaohan Liu, Hui Liu, Jiwen Liu, Xiaoqiang Liu, Yuan Lu, Shun Mao, Yongsen Shao, Yingchao Shi, Huafeng Shi, Xiaoyu Sun, Peiqin Tang, Songlin Wan, Pengfei Wang, Chao Wang, Xuebo Zhang, Haoxian Zhang, Yuanxing Zhou, Yan |
| author_facet | Kling Team Chen, Jialu Ding, Yikang Fang, Zhixue Gai, Kun Gao, Yuan He, Kang Hua, Jingyun Jiang, Boyuan Lao, Mingming Li, Xiaohan Liu, Hui Liu, Jiwen Liu, Xiaoqiang Liu, Yuan Lu, Shun Mao, Yongsen Shao, Yingchao Shi, Huafeng Shi, Xiaoyu Sun, Peiqin Tang, Songlin Wan, Pengfei Wang, Chao Wang, Xuebo Zhang, Haoxian Zhang, Yuanxing Zhou, Yan |
| contents | Avatar video generation models have achieved remarkable progress in recent years. However, prior work exhibits limited efficiency in generating long-duration high-resolution videos, suffering from temporal drifting, quality degradation, and weak prompt following as video length increases. To address these challenges, we propose KlingAvatar 2.0, a spatio-temporal cascade framework that performs upscaling in both spatial resolution and temporal dimension. The framework first generates low-resolution blueprint video keyframes that capture global semantics and motion, and then refines them into high-resolution, temporally coherent sub-clips using a first-last frame strategy, while retaining smooth temporal transitions in long-form videos. To enhance cross-modal instruction fusion and alignment in extended videos, we introduce a Co-Reasoning Director composed of three modality-specific large language model (LLM) experts. These experts reason about modality priorities and infer underlying user intent, converting inputs into detailed storylines through multi-turn dialogue. A Negative Director further refines negative prompts to improve instruction alignment. Building on these components, we extend the framework to support ID-specific multi-character control. Extensive experiments demonstrate that our model effectively addresses the challenges of efficient, multimodally aligned long-form high-resolution video generation, delivering enhanced visual clarity, realistic lip-teeth rendering with accurate lip synchronization, strong identity preservation, and coherent multimodal instruction following. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_13313 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | KlingAvatar 2.0 Technical Report Kling Team Chen, Jialu Ding, Yikang Fang, Zhixue Gai, Kun Gao, Yuan He, Kang Hua, Jingyun Jiang, Boyuan Lao, Mingming Li, Xiaohan Liu, Hui Liu, Jiwen Liu, Xiaoqiang Liu, Yuan Lu, Shun Mao, Yongsen Shao, Yingchao Shi, Huafeng Shi, Xiaoyu Sun, Peiqin Tang, Songlin Wan, Pengfei Wang, Chao Wang, Xuebo Zhang, Haoxian Zhang, Yuanxing Zhou, Yan Computer Vision and Pattern Recognition Avatar video generation models have achieved remarkable progress in recent years. However, prior work exhibits limited efficiency in generating long-duration high-resolution videos, suffering from temporal drifting, quality degradation, and weak prompt following as video length increases. To address these challenges, we propose KlingAvatar 2.0, a spatio-temporal cascade framework that performs upscaling in both spatial resolution and temporal dimension. The framework first generates low-resolution blueprint video keyframes that capture global semantics and motion, and then refines them into high-resolution, temporally coherent sub-clips using a first-last frame strategy, while retaining smooth temporal transitions in long-form videos. To enhance cross-modal instruction fusion and alignment in extended videos, we introduce a Co-Reasoning Director composed of three modality-specific large language model (LLM) experts. These experts reason about modality priorities and infer underlying user intent, converting inputs into detailed storylines through multi-turn dialogue. A Negative Director further refines negative prompts to improve instruction alignment. Building on these components, we extend the framework to support ID-specific multi-character control. Extensive experiments demonstrate that our model effectively addresses the challenges of efficient, multimodally aligned long-form high-resolution video generation, delivering enhanced visual clarity, realistic lip-teeth rendering with accurate lip synchronization, strong identity preservation, and coherent multimodal instruction following. |
| title | KlingAvatar 2.0 Technical Report |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2512.13313 |