Google’s artificial intelligence research lab Deep mindhas announced a new tool that voices AI-generated videos called V2A (short for “audio from video”). According to DeepMind’s statement V2AWhen creating the sounds, both the written description and the video are analyzed pixel by pixel. This way it can create sound even if you don’t write a description.

It must be said that artificial intelligence tools are important in dubbing the video. Because while there are many artificial intelligence tools today that can create videos from text or images, these tools are It can produce mainly silent videos. However, we all know that video is a type of media that is both audio and visual. When we see a video of a fast moving train, the video loses its effect if we cannot hear the sound of the rails, wagons and locomotive.

Example of V2A voicing an AI-generated guitar playing video

DeepMind claims that V2A is unbeatable when it comes to producing video-compatible audio and automatically syncing audio with the image. However, looking at the examples shared, it’s hard to say we’re impressed. V2Aproduces sound to match the picture, but is more like dubbing a standard sound to match the picture.

The examples below show how far V2A has come when it comes to video dubbing. primitive It can help you understand what’s left. However, we shouldn’t forget that Dall-E once drew a fluffy yellow ball when asked to draw a canary. Artificial intelligence tools improve as you use them…