Microsoft Unveils VASA-1, An Eerily Realistic AI That Animates Still Portraits

Microsoft (Nasdaq: MSFT) has announced VASA-1, an AI system capable of bringing still images to life. The model uses a single photo of a face and an audio clip to generate a stunningly realistic video of that person speaking the audio.

VASA-1 meticulously animates facial expressions, lip movements, and even head motions to create an utterly convincing illusion of the subject talking or singing.

The results are so lifelike that an inanimate 16th-century Mona Lisa painting can be made to rap with modern inflections.

While the potential applications span digital avatars, accessibility aids, and educational tools, experts warn the technology could enable disturbing deepfake forgeries. Scammers might impersonate trusted individuals to defraud people through fabricated video messages.

“Seeing is most definitely not believing anymore,” cautioned security researcher Jake Moore told MailOnline. “As this technology improves, people should think twice before accepting correspondence as genuine.”

Selkirk Copper Mines — sponsored Sponsored · Selkirk Copper Mines

Related: Almost 6 Months In, Is NVIDIA’s Eye Contact AI Still Too Creepy To Use?

Microsoft researchers claim the system is solely intended for positive use cases that avoid deception. However, they acknowledge that “it could still potentially be misused for impersonating humans.” The team is exploring ways to apply VASA-1 for detecting deepfake videos instead.

Related: Grimes Offers to Split Royalties for AI ‘Collabs’

The researchers note that generated videos still contain “identifiable artifacts” distinguishing them from reality for now. But the incredible photorealism achieved raises alarms about a future when AI-synthesized media becomes indistinguishable from authentic recordings.

“Given such context, we have no plans to release an online demo, API, product, additional implementation details, or any related offerings until we are certain that the technology will be used responsibly and in accordance with proper regulations,” the research team wrote.


Information for this story was found via Microsoft, MailOnline, and the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.

Video Articles

A $2.2B Gold Project Is Outgrowing Its Plan | Michael Henrichsen – Gold X2 Mining

Pay for the Copper, Get the Gold Free | Rob McEwen – McEwen Inc

This Gold Discovery Was Already Huge. Now It’s Becoming a Monster. | Goliath Resources

Recommended

This Gold Discovery Just Keeps Getting Bigger | Roger Rosmus – Goliath Resources

Cardiol Expands MAVERIC Trial to Up to 150 Patients After Reaching Enrollment Target Early

Related News

Sam Altman Wants To Raise Trillions For The Future Of AI

OpenAI CEO Sam Altman wants to transform the global semiconductor industry. The catch: he needs...

Friday, February 9, 2024, 10:38:00 AM

Reddit Signs $60 Million Per Year Deal For AI Training Data

$60 million per year, so far — that’s how much licensing Reddit’s vast trove of...

Tuesday, February 20, 2024, 07:40:07 AM

An Old Profile Reveals Sam Altman Has Guns and IDF Gas Masks In Case AI Takes Over

Sam Altman, CEO of the artificial intelligence startup OpenAI, has always wanted to save the...

Wednesday, November 29, 2023, 06:44:00 AM

Amazon to Acquire AI Chipmaker Perceive for $80 Million

Amazon (Nasdaq: AMZN) has announced plans to acquire Perceive Corp, an AI chipmaker and model...

Wednesday, August 21, 2024, 12:29:54 PM

This New Video-Based AI-Powered Tool Lets You ‘Speak’ In Another Language

HeyGen, a generative AI video program, has unveiled a new feature that allows users to...

Wednesday, September 13, 2023, 12:46:00 PM