DeepFaceLab: A tool that utilizes ML to replace faces in videos

I wonder what this implies for the future of media (audio, photos, video) in general? Do we only consume media that has been signed by the creator and verified by an authority that we trust, e.g., a "blue checkmark" for media? Do we know how effective the SSL certificate verification has been in browsers at influencing consumer behavior?

Video evidence is going to be inadmissible very soon.<p>It&#x27;s ironic that we&#x27;re essentially being pushed back to a pre-technology, pre-media time. &quot;If you didn&#x27;t see it with your own two eyes, you can&#x27;t believe it&quot;

I think the cons are going to heavily outweigh the pros with this tech.<p>Fake news on high octane race gas.

There is also:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;deepfakes&#x2F;faceswap" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;deepfakes&#x2F;faceswap</a><p>With a big development community and interesting results.

What are the non-malicious uses for something like this? Haven&#x27;t actually got around to pondering that yet.

Is it the same technology used by Zao? The example in the gallery does not look as good as the video here <a href="https:&#x2F;&#x2F;twitter.com&#x2F;AllanXia&#x2F;status&#x2F;1168049059413643265" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;AllanXia&#x2F;status&#x2F;1168049059413643265</a>

Hollywood has been doing this for a while, for stunt performers. Sometimes well, sometimes badly. We&#x27;ll be seeing more of that as it works better.<p>As for the political implications, go watch &quot;Wag the Dog&quot; again.

This technology has been posted on HN before and the comments were all too similar to this one. I understand why a lot of people are falling into the FUD-hole, but we can pretty easily solve the issues mentioned in the comments by inserting a digital key - similar to a SSL certificate, into an image&#x2F;video upon initial upload. Or even upon creation, maybe using some sort of ledger to verify integrity. I&#x27;m not too worried.

Am I the only one that cries when seeing non PEP8 compliant Python code? Why do data scientists mistreat Python that way?

On a (somewhat) related note, I&#x27;m working on face recognition with homomorphic encryption, therefore without compromising the user privacy. The bold goal is the first privacy preserving videocamera. If you find this interesting, I would love to chat about it.

I&#x27;m surprised no one&#x27;s bringing up the precedent of photoshop. For years we&#x27;ve culturally realized that pictures may not tell the whole story, and this is the exact same, down to results that can&#x27;t quite escape the uncanny valley.

On a scale of 1 to 10, how much easier does this make Deepfakery?

Deep Fake with its revenge porn wasn&#x27;t enough to give this a second thought?