Google Is Making It Easy To Deepfake Yourself

Google's new AI video model allows you

to create deep fakes of yourself.

At Google I/O, the company announced Omni Flash,

its new video generation model

that'll power everything Google.

The big pitch is that Omni Flash makes AI video look better.

More detail, more consistency,

fewer warped faces between generations.

But the feature that I can't stop thinking about

is much creepier.

Google calls it avatars.

Basically, you can scan your face, capture your voice,

and then generate videos starring an AI version of yourself,

which means that you can take a video,

well, generate a video,

that looks like you are on camera

without ever actually appearing on camera again.

Google says this is for creators

who wanna bring themselves into new content

without having to actually shoot that new content,

which sounds convenient, but also is a little dystopian,

and this is where the similarities start popping up.

OpenAI's Sora app had a similar selfie-cloning feature

where users can make AI videos starring themselves,

but Sora was an AI-first social network.

Depending on your settings,

other people could also generate videos of you.

Google's version is different, well, at least for now.

The company says its initial focus is on letting users

generate AI videos of themselves only, not other people.

And instead of living inside of a standalone app,

these avatars are being added

to Google's broader creation tools,

whether it's Flow, Gemini, or YouTube.

And that matters, because Sora felt like this

strange experimental social app.

Google's version feels more like this technology

is being folded into the mainstream creator internet.

To set it up, users can scan a QR code

and record themselves saying a string of numbers,

moving their head around

so Google can capture every angle of your pretty face.

The company says that videos made with Omni,

including these avatar videos,

will include its watermark, called SynthID,

that marks these as AI-generated.

But even with those safeguards,

the direction here is pretty clear.

The next phase of AI video is not just typing a prompt

and getting a fake scene,

it's putting yourself inside of that fake scene.

These tools can make content production faster and easier,

but they also push us closer to an internet

where a video selfie might not actually be a video selfie.