T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908541669867520 |
|---|---|
| author | Li, Jiachen Long, Qian Zheng, Jian Gao, Xiaofeng Piramuthu, Robinson Chen, Wenhu Wang, William Yang |
| author_facet | Li, Jiachen Long, Qian Zheng, Jian Gao, Xiaofeng Piramuthu, Robinson Chen, Wenhu Wang, William Yang |
| contents | In this paper, we focus on enhancing a diffusion-based text-to-video (T2V) model during the post-training phase by distilling a highly capable consistency model from a pretrained T2V model. Our proposed method, T2V-Turbo-v2, introduces a significant advancement by integrating various supervision signals, including high-quality training data, reward model feedback, and conditional guidance, into the consistency distillation process. Through comprehensive ablation studies, we highlight the crucial importance of tailoring datasets to specific learning objectives and the effectiveness of learning from diverse reward models for enhancing both the visual quality and text-video alignment. Additionally, we highlight the vast design space of conditional guidance strategies, which centers on designing an effective energy function to augment the teacher ODE solver. We demonstrate the potential of this approach by extracting motion guidance from the training datasets and incorporating it into the ODE solver, showcasing its effectiveness in improving the motion quality of the generated videos with the improved motion-related metrics from VBench and T2V-CompBench. Empirically, our T2V-Turbo-v2 establishes a new state-of-the-art result on VBench, with a Total score of 85.13, surpassing proprietary systems such as Gen-3 and Kling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_05677 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design Li, Jiachen Long, Qian Zheng, Jian Gao, Xiaofeng Piramuthu, Robinson Chen, Wenhu Wang, William Yang Computer Vision and Pattern Recognition Artificial Intelligence In this paper, we focus on enhancing a diffusion-based text-to-video (T2V) model during the post-training phase by distilling a highly capable consistency model from a pretrained T2V model. Our proposed method, T2V-Turbo-v2, introduces a significant advancement by integrating various supervision signals, including high-quality training data, reward model feedback, and conditional guidance, into the consistency distillation process. Through comprehensive ablation studies, we highlight the crucial importance of tailoring datasets to specific learning objectives and the effectiveness of learning from diverse reward models for enhancing both the visual quality and text-video alignment. Additionally, we highlight the vast design space of conditional guidance strategies, which centers on designing an effective energy function to augment the teacher ODE solver. We demonstrate the potential of this approach by extracting motion guidance from the training datasets and incorporating it into the ODE solver, showcasing its effectiveness in improving the motion quality of the generated videos with the improved motion-related metrics from VBench and T2V-CompBench. Empirically, our T2V-Turbo-v2 establishes a new state-of-the-art result on VBench, with a Total score of 85.13, surpassing proprietary systems such as Gen-3 and Kling. |
| title | T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2410.05677 |