Latent-Shift: Latent Diffusion with Temporal Shift

We propose Latent-Shift -- an efficient text-to-video generation method based on a pretrained text-to-image generation model that consists of an autoencoder and a U-Net diffusion model. Learning a video diffusion model in the latent space is much more efficient than in the pixel space. The latter is often limited to first generating a low-resolution video followed by a sequence of frame interpolation and super-resolution models, which makes the entire pipeline very complex and computationally expensive. To extend a U-Net from image generation to video generation, prior work proposes to add additional modules like 1D temporal convolution and/or temporal attention layers. In contrast, we propose a parameter-free temporal shift module that can leverage the spatial U-Net as is for video generation. We achieve this by shifting two portions of the feature map channels forward and backward along the temporal dimension. The shifted features of the current frame thus receive the features from the previous and the subsequent frames, enabling motion learning without additional parameters. We show that Latent-Shift achieves comparable or better results while being significantly more efficient. Moreover, Latent-Shift can generate images despite being finetuned for T2V generation. We present the first method capable of photorealistically reconstructing a non-rigidly deforming scene using photos/videos captured casually from mobile phones.

An illustration of our framework. From left to right: (a) An autoencoder is trained on images to learn latent representation. (b) The pretrained autoencoder is adapted to encode and decode video frames independently. (c) During training, a temporal shift U-Net $\epsilon_\theta$ learns to denoise latent video representation at a uniformly sampled diffusion step $t\in [1,T]$. During inference, the U-Net gradually denoises from a normal distribution from step $\widehat{T}-1$ to $0$ where $\widehat{T}$ is the number of resampled diffusion steps in inference. (d) The U-Net $\epsilon_\theta$ is composed of two key building blocks: the $2$D ResNet blocks with convolutional layers, highlighted in violet, and the transformer blocks with spatial attention layers, colored in gray. The temporal shift module, highlighted in red, shifts the feature maps along the temporal dimension. It is inserted into the residual branch of each $2$D ResNet block. The text condition is applied to the transformer blocks via cross-attention. The channel dimension $c$ in the latent space representation of $\mathbf{z}$ and $\mathbf{u}$ are omitted for clarity.