I've been reading that this kind of thing will be used in political videos of politicians like to swing elections or favor, or geopolitical things
Since I read that years ago, I think its more likely that we will simply get a bombshell article a few years in the future that its been in use already for a very long time! Similar to how Cambridge Analytica information and other information warping info came out years after it was already in play on a broad scale.
Not sure about open-source, but a number of friends of mine who work as freelance professional creatives (e.g. ad hoc video advertisements for smaller companies) espouse Topaz Labs’ Video Enhance AI [0] for this sort of work. Anecdotally, seeing the results of their applications of it, it does seem to work astoundingly well— especially for an application that runs fine on hobbyist/enthusiast grade hardware (albeit the render times will likely still be in a multiple of hours for longer videos; personal experience on 3090 / 5950X / 128GB home work machine was roughly 5-8 frames rendered/sec when upscaling 720p footage to 1440p, roughly 12 seconds render time for every second of footage. I imagine the render time would scale with the demands of the selected AI model / output resolution).
I am not aware of any singular FOSS project in creative work that performs similarly to Topaz Labs’ product for all input videos (note to reader: if I am wrong, correct me— I only have second-hand anecdotes to go off of from my friends who work with this kind of work in their careers, and precious little first-hand experience from upscaling old family memories to experiment with the software). As far as I am aware, this is because different upscaling models are trained on, and thus effective / mainly used for a specific type of content. The same upscaling model that was trained on interpolated 480p videos with compression artifacts will not produce the same results as one trained on, say, anime videos/manga, e.g. the models used by Waifu2X [1]. Hence, with Topaz Labs’ application, you select the model that was trained on footage that best matches the footage you wish to upscale.
All that being said, I do know that some (if not all) of the upscaling models Topaz uses are FOSS. Much of their application is just syntactical sugar upon the models it uses, making it easier for non-SWEs to use. I’m not sure whether or not the models that Topaz distributes have any level of training done by them, in house— logically, I would assume as such, otherwise their product wouldn’t be as performant as it is.
Not cutting edge, but there are various deep learning upscaling tools for individual photos on GitHub. You could unpack a video into audio streams and individual frames with ffmpeg on a GPU cloud instance, upscale all the individual frames, and then repack it with ffmpeg.
However, the examples I saw hit the uncanny valley for me. The changes are realistic but there is just something off about them, that bothers my brain.
I think that in the future, the main thing that will identify a real video from a fake will be the reputation of the publisher. Same as for photos and quotes now.
It's as trivial to sign fake content as real one, and people are really, really, really bad at anything close to due diligence for signature verification and revocation (as that's an objectively hard problem).
Furthermore, there's the catering to the least denominator. Whatever signing ability is available for a random widespread cheap third world smartphone used in Tik Tok videos will be treated as sufficiently good to assume it to be true; so if a determined attacker wanting to create a misinformation campaign with fake videos can circumvent the security of that signing process (e.g. get a bunch of valid keys indistinguishable from these phones, then fake videos will have as good signing as real videos.
We are talking about the future here. Easy to use certificate authorities would be available if it came to this. It would be part of the software ecosystem, part of mobile operating systems for example. Your cert would be registered with google and, for example, a camera app would use an OS level API to sign an image when it is taken. Anyone who wants to verify that it's your image could check against your public key registered with google or whatever well known, secure, trusted service.
If a random poor third world person can easily obtain a valid certificate, then any attacker can also do so - identity theft or totally fake identities are a thing despite all our best efforts and it's naive to assume that this will be magically solved in the future; and if it is not easy a random poor third world person to do so, then they won't, and all their real, valid user-generated content (e.g. real cell phone videos from conflict zones) won't be signed, and the society will consider unsigned videos as valid.
Furthermore, any attacker with the desire and resources to create a disinformation campaign can simply recruit a new real person to sign each deepfake campaign, just as criminals now hire money laundering mules (I mean, that expense would be less than the actual effort of creating the media) and any intelligence agency using deepfakes for propaganda can literally create new valid identities by fiat (just issue new real passports/birth certificates/whatever for nonexistent people) that are indistinguishable from real people as far as google or anyone else abroad can verify.
In essence, what you describe would work if and only if we had a global, trusted database of all people worldwide that doesn't allow for fake people. We're very far from that, and the obstacles for that aren't technological, it's definitely not something that Google can solve.
I get the feeling that in the future, it will be necessary to sign videos with cryptographic keys based off of scans of your iris or other secure biomarkers if you want anyone to know it's actually real. Mannerisms are much too imitable ;)
What could possibly prevent me from signing a fake video with a cryptographic key based off of scan of your biomarker, once I obtain that scan somehow ? Biomarkers can't be changed or revoked, so once they leak, you either have to change the whole system or stop trusting your content forever.
(the details of obtaining the marker aren't relevant, I think you'll agree that one way or another there will be some leaks or cracked phones or any one of other possibilities eventually exposing biomarkers for at least some of the population)
