News

An experimental AI device from Microsoft can turn photos into videos

Jakarta (MidLand) – Microsoft Research Asia introduces artificial intelligence tools (artificial intelligence/AI) experimental called VASA-1 which can take photos or still images and existing audio files to create some sort of video.

According to the post Dedicate onself on Saturday (4/20), the AI ​​model was able to create videos showing people talking by including photos and voice samples.

VASA-1 reportedly has the ability to produce facial expressions and head movements from photos of people, as well as lip movements that match the audio of conversations or songs.

The researchers have posted many examples on the project page and the results look so good that they might make people believe they are real.

While the head and lip movements in the examples still appear robotic and out of sync when viewed up close, the technology could easily be misused to create fake videos of a person.

The researchers were aware of the potential dangers and decided not to release “online demo tools, APIs, products, further implementation details, and other related matters” until they were confident that their technology “will be used responsibly and in accordance with the regulations”.

Read also: Microsoft to invest IDR 46 trillion in Japan for AI data centers
Read also: Qualcomm announces new generative AI for devices

The development team believes that VASA-1 offers many benefits despite the potential for abuse.

They say this technology can be used to increase educational equity, increase accessibility for people who have communication difficulties, and provide a conversation partner and therapeutic support for those who need it.

According to scientific publications on the technology, VASA-1 was trained using the VoxCeleb2 dataset containing more than one million utterances for 6,112 celebrities taken from YouTube videos.

Although trained using faces of real characters, VASA-1 can also work on artistic images such as the Mona Lisa, which the researchers combined with audio files of actress Anne Hathaway.

Read also: Humane releases Ai Pin, a wearable artificial intelligence device
Read also: Intel reveals first AI processor called Springhill

Translator: Farhan Arda Nugraha
Publisher: Maryati
Copyright © MidLand 2024

Quoted From Many Source

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button