Phenaki_云技派,专注ai人工智能! 124

What is Phenaki?

Phenaki is a project from Google Research whose goal is to synthesize realistic videos from open-domain text descriptions. It can generate videos of several minutes in length directly from text.

With Phenaki, you can also generate videos from still images and prompts. The proposed video encoder-decoder outperforms all per-frame baselines currently used in the literature in terms of spatiotemporal quality and the number of tokens per video.

To generate video tokens from text, Phenaki uses a bidirectional masked transformer conditioned on pre-computed text tokens. The generated video tokens are subsequently de-tokenized to create the actual video.

Phenaki is a breakthrough in the field of video synthesis because it can handle open-domain and temporally variable prompts, unlike previous methods that were limited by data availability and computational cost. Phenaki can also generate videos from still images and prompts, such as zooming into a cat's eyes or making it yawn. Phenaki has many potential applications, such as entertainment, education, storytelling, and art.

What can Phenaki do?

This video went viral online, Google made AI video forgery too real and too easy

Phenaki Advantages:

It can generate videos of variable length and quality, up to two minutes long

It can handle open-domain and temporally variable prompts, such as stories or descriptions

It leverages a large corpus of image-text pairs and a small number of video-text examples to generalize beyond the video dataset