Appreciate the video Uni & it's nice looking too btw but even videos are becoming easily faked today. Machine learning (subset of A.I.) is giving us software that makes it so even a basic user can mock up high quality video fakes.
Take a look at technology called
Deepfake. Some software is so simple you literally just add words or phrases, inserting a few headshot snaps, then using machine learning algorithms it superimposes/morphs to create a very accurate fake.
Example of Kate McKinnon impersonating Hillary Clinton on SNL during 2016 presidential debates but with the real Hillary Clinton swapped in using Kate's dialogue via Deepfake technology:
Link to Deepfake of real Hillary Clinton (on right) swapped with Kates dialogue from SNL skit (on left) showing just how compelling even nascent ML technology can look today.
Note: I linked it to play @
2:12 but it's a WatchMojo list of varying types of Deepfake technologies + top 10 list of techniques/examples.
To be fair, yes, this may still be a bit difficult for the novice person, however, tools are available that do make some of it as easy as inserting few images, face(s), text/dialogue, let it process, voila! Some are literally that simple. Videos are still harder to fake, yep indeed, but my main point is that today there really isn't any 100% absolutely fail safe way to know other than corroboration from the community, seeing yourself, ask for real time Skype/Facetime (if parties are okay with it), or just trust them then leave if it's not what you saw (goes for man or woman, and report it in your private locker, or publicly; again community corroboration). Or simply use techniques like images with dates, today's weather, detailed items that are
"time boxed" and would be hard to photoshop, etc... The average user is not going to be able to photoshop pics like that with any level of believability.
Though it is quickly becoming easier to fake a video using some new tools due to simply needing a few face snaps and/or inserting some dialogue and the software does it for you versus the picture option b/c then the individual would need to do that themselves.
Faking a video using programmatic techniques is easier b/c it's bound by a fairly rigid problem/concept domain (facial features + potentially dialogue; it's even easier if there is no dialogue, very easy).
Pictures can contain numerous items that are used to identify when it was taken, special items that couldn't be in the picture unless the person just took it
(i.e. "time boxed" material - today's weather, unique items only the person would know or could request 3min ago, etc...). It is much harder to use algorithms to fake that b/c you need a human to select the items that need editing and exactly
what needs editing (problem domain is infinitely larger).
Just something for everyone to be aware of and more food for thought