A new era of deepfake videos + Instagram AI Influencer coming?

Microsoft Researchers show off VASA-1 - A tool to create deepfakes from a single picture and audio

Microsoft's new AI model, VASA-1, can create realistic videos of a person talking or singing from just one photo and an audio clip.

The videos seem to be rendered almost instantaneously and are impressively photorealistic.

These Audio2Video models become almost indistinguishable from actual recordings.

Just two months ago Alibaba Researchers introduced EMO, which is similar but doesn’t produce as polished facial expressions.

What does this mean?

While VASA-1 is not publicly available (and probably won’t be), its a sneak preview into what will soon be ubiquitous technology.

Any 2nd grader will be able to clone a voice from a short voice recording, take a high resolution picture and create a photorealistic deepfake video.

Meta’s Creator A.I. is a tool for Instagram influencer to engage with fans using chatbots

Instagram is testing a new AI program called "Creator AI" to help influencers manage interactions with fans more efficiently (in Beta).

This program will use chatbots that replicate the influencers style of communication to respond to direct messages and, possibly later, comments.

Meta claims that all AI responses will have proper disclosures, so fans are aware that they are communicating with an AI. Source

Is AI enhancing or replacing Influencer?

Meta claims that the Creator A.I. program is part of Meta's broader push into AI across its services and help to manage large volumes of fan interactions.

However, it might also be the first step into replacing the physical influencer with an AI representation and leveraging AI avatars to create individual video content or live avatar interactions.

Celebrities leveraging their fame through AI representation is a trend that has also reached Hollywood, with one of the biggest talent agencies now testing to license AI avatars of their talent.

Microsoft rolls out “Content Integrity” tools using C2PA to support global elections

As a founding member of the Content Authenticity Initiative, Microsoft introduced a set of software tools to allow organisations to authenticate their media content.

This includes verifying the origin and alterations of images, videos, and audio files attaching secure "Content Credentials”, which uses cryptographic signatures using the C2PA standard.

Microsoft emphasizes the importance of these tools in maintaining healthy democracies by ensuring media integrity.

The tools are being provided to EU political parties, campaigns, and news organizations in a private preview at no cost throughout 2024. Source

Is this the solution?

The C2PA standard is (currently) the widest adopted initiative to create a common standard in content provenance, which is a tamper proof way of authenticating the origin and authenticity of a media file.

Even though many GenAI platforms like Adobe’s Firefly, OpenAI’s Dall-E and many more already attach c2pa standardised metadata to any created content, it still lacks any adoption of media publishers and social media platforms to display the information (the Content Credentials) and properly label content.