Abstract: Provided is a video generation model for performing text-to-video (T2V) or other video generation techniques. The proposed model reduces the computational costs associated with video generation. In particular, unlike traditional T2V methods, the disclosed technology can generate the full temporal duration of a video clip at once, bypassing the need for extensive computation. As one example, a machine-learned denoising diffusion model can simultaneously process a plurality of noisy inputs that correspond to various timestamps spanning the temporal dimension of a video to simultaneously generate synthetic frames for the video that match the timestamps.

Abstract: Using a pre-trained and fixed Vision Transformer (ViT) model as an external semantic prior, a generator is trained given only a single structure/appearance image pair as input. Given two input images, a source structure image and a target appearance image, a new image is generated by the generator in which the structure of the source image is preserved, while the visual appearance of the target image is transferred in a semantically aware manner, so that objects in the structure image are “painted” with the visual appearance of semantically related objects in the appearance image. A self-supervised, pre-trained ViT model, such as a DINO-VIT model, is leveraged as an external semantic prior, allowing for training of the generator only on a single input image pair, without any additional information (e.g., segmentation/correspondences), and without adversarial training. The method may generate high quality results in high resolution (e.g., HD).

Abstract: Using a pre-trained and fixed Vision Transformer (ViT) model as an external semantic prior, a generator is trained given only a single structure/appearance image pair as input. Given two input images, a source structure image and a target appearance image, a new image is generated by the generator in which the structure of the source image is preserved, while the visual appearance of the target image is transferred in a semantically aware manner, so that objects in the structure image are “painted” with the visual appearance of semantically related objects in the appearance image. A self-supervised, pre-trained ViT model, such as a DINO-VIT model, is leveraged as an external semantic prior, allowing for training of the generator only on a single input image pair, without any additional information (e.g., segmentation/correspondences), and without adversarial training. The method may generate high quality results in high resolution (e.g., HD).