Google Unveils VideoPoet AI That Generates Short Videos from Text
-
Google VideoPoet is an experimental AI system that can generate short videos from text prompts. It was trained on over 1 billion image-text pairs and 270 million videos.
-
VideoPoet works differently than other text-to-image AIs. It directly predicts video and audio chunks like a language model predicts words.
-
VideoPoet can generate longer and more complex videos with motion than competing models. The videos currently top out at 8-10 seconds.
-
VideoPoet has additional capabilities beyond text-to-video generation like applying styles and effects to existing videos.
-
Google hasn't announced availability details for VideoPoet yet. It will likely require a subscription like other AI image/video generators.