Filtered to generative models · clear filter
This paper creates a new AI model that can edit and generate videos without needing masks, and can also learn to mimic image editing capabilities. Practitioners might care about this because it could lead to more diverse and realistic video editing data, and enable AI models to understand and generate human-like video editing instructions.
This paper proposes a new training method for autoregressive video diffusion models called Self Gradient Forcing, which helps them better remember and use past information to generate future frames. Practitioners might care about this because it could lead to more realistic and stable video extrapolation.
This paper introduces Oxygen-TryOn, a new AI model that can generate photorealistic images of people wearing any fashion item, in any setting. Practitioners in the fashion industry might care about this model because it can revolutionize virtual try-on, allowing for more realistic and diverse scenarios.
This paper introduces Wonder, a system that can generate videos of a virtual world in real-time, allowing users to interact with and explore the environment. Practitioners in fields like video game development or virtual reality might care about this technology for its potential to create more immersive experiences.
This paper introduces Explorative Modeling, a new approach to training generative models that allows for end-to-end generation by exploring multiple candidate matches between model generations and data. This can lead to improved performance and efficiency in various applications.
This paper introduces Three-Body Scattering Modeling (TBSM), a new approach to generative modeling that uses a distributional energy to induce motion and provide direct regression supervision for one-step generators. Practitioners might care because TBSM can achieve high-quality generation with reduced field noise and fewer sample-level losses.
This paper introduces Parallel Decoding Distillation (PDD), a new method to speed up image and video generation in diffusion and flow models, achieving state-of-the-art performance while improving video diversity. Practitioners might care about PDD's potential to accelerate video generation tasks.
This paper introduces Chamaileon, a new method for designing protein binders that can model multiple targets and conformations, which could help researchers and practitioners design more advanced proteins with specific functions. Practitioners might care about this approach because it could lead to breakthroughs in structural biology and protein engineering.