Phenaki: Variable Length Video Generation From Open Domain Textual Description

TL;DR

Phenaki is a novel model that enables realistic, variable-length video generation from open domain text prompts by using a discrete token-based representation and joint training on image and video datasets.

Contribution

It introduces a new video representation learning method with a causal attention tokenizer and a bidirectional transformer conditioned on text, allowing for flexible, long video generation from prompts.

Findings

Generates arbitrary long videos conditioned on prompt sequences

Achieves better spatio-temporal consistency than per-frame baselines

Generalizes beyond available video datasets through joint training

Abstract

We present Phenaki, a model capable of realistic video synthesis, given a sequence of textual prompts. Generating videos from text is particularly challenging due to the computational cost, limited quantities of high quality text-video data and variable length of videos. To address these issues, we introduce a new model for learning video representation which compresses the video to a small representation of discrete tokens. This tokenizer uses causal attention in time, which allows it to work with variable-length videos. To generate video tokens from text we are using a bidirectional masked transformer conditioned on pre-computed text tokens. The generated video tokens are subsequently de-tokenized to create the actual video. To address data issues, we demonstrate how joint training on a large corpus of image-text pairs as well as a smaller number of video-text examples can result in generalization beyond what is available in the video datasets. Compared to the previous video generation methods, Phenaki can generate arbitrary long videos conditioned on a sequence of prompts (i.e. time variable text or a story) in open domain. To the best of our knowledge, this is the first time a paper studies generating videos from time variable prompts. In addition, compared to the per-frame baselines, the proposed video encoder-decoder computes fewer tokens per video but results in better spatio-temporal consistency.