I vaguely recall reading a similar thread but I think it was in IMHO and i’m looking for, well, the straightdope.
Every time a slightly unusual video is posted the cries of FAKE! go up, even when its pretty obvious its not actually faked. And there is an idea that a single person can fairly easily create a faked video although all the actually faked videos shown required a team of people and significent time, effort and expenditure to come across as plausibly real.
I’m not talking about re-cutting or dubbing extant footage but making significent changes to footage or coming up with entirely new footage that doesn’t show what it appears to show (I’m thinking of things like the footage of a bear chasing a skier, it was convincing but was also not something the stereotypical lone nerd in a basement could have put together).
Basically I don’t think its as easy to convincingly produced faked video footage as it is commonly believed to be, though I’m perfectly willing to be proven wrong.
Most of the hardware and software used by pros isn’t that different from what lone basement nerds can get ahold of. Someone with the time and talent should be able to fake a great deal, as long as it isn’t trying to fake people, which still is non-trivial. If you mean compositing two video sources, it would be only tedious, not difficult–you might have to tweak every frame of the bear (for example) by hand to make sure no pixels from the original background are present and no pixels from the bear are absent. You’d have to hope the lighting in both sources was from the same direction, but even that could be potentially tweaked with even more time spent. You’d have to hand-draw new shadows. I could do it, if I was willing to put tens of hours of work into a few seconds of video.
There’s a really amazing video series by a Youtuber called CaptainDisillusion who takes these fake viral videos and breaks them down to show you how people did them, in most cases using very simple techniques both physical and computer-visual.
Basically there’s a lot of programs out there that actually do most of the work for you, it’s up to the uploader to hide the seams.
Not too long ago I read a news article where there is a research group actually being paid to work on this problem. Kinda wish I had bumped my thread with it now.
But with today’s video cards and software, CGI isn’t exclusive to Pixar and Dreamworks.
The mention of You Tube made me think you were asking about the ability to claim to have a popular video on You Tube and show someone the evidence of that such they’d believe that, while not actually having such video or at least not having it be as popular as you claim it is. You are actually talking about faking a video that has absolutely nothing to do with YouTube at all other than that you plan on hosting it there, because that’s what everyone else does. I suppose it’s gotten to the point that absolutely every video ever made is available on YouTube, and thus using “YouTube” before video is simply redundant and part of a force of habit, and not actually descriptive of the video at all.
Some times no CGI is needed; just simple misdirection. For example, a few years back there were some videos that showed how you could supposedly unlock a car door with a tennis ball. A lot of people tried this technique and couldn’t repeat the success they had seen in the video. The reason was that the people in the videos had somebody standing just off camera who would unlock the car door with a key remote as the tennis ball was being applied to the door.
I follow a couple of video oriented forums while I don’t know anything about editing videos, I see two common questions are “How do I do this effect” and “How do I add / remove a person/object from a video”. “How to do this effect”, is often built-into modern video editors or as an add-on filter. The answer for “How do I add/remove a person/object from a video” is rotoscoping. Basically a cut and paste for video. Again, what would take hours or days previously is nearing realtime with today’s PCs.
One of the current big issues for celebrities is fake porn where their face is superimposed over someone else’s body. It’s reportedly not truly realistic yet, but in another decade or less, it will be impossible to prove otherwise unless the original video is found.
Not really–rotoscoping is a specific thing–drawing (or rendering) animation on top of other video, and it has been around for a long time. It isn’t the term for removing someone from a video, and isn’t the term for adding another live element over a video (that is called compositing.) Rotoscoping is fairly easy because you draw or render only what you want and then place that over the background. Compositing is harder–yes, there are “magic wand” tools that attempt to outline and cut out a section of a photo or video, but it isn’t simple point-and-click–unless the object you want to cut out is at a very high and clear contrast against the background (such as standing in front of a blue/green screen) there is going to be some halo around the object or some pixels of the object chipped away–if you want to make even a fairly decent fake, you’ll need to clean the edges of the object by hand–there may be a time when the software is as good as the human eye at picking out an object against a naturally-noisy background, but it isn’t 2018.
As tedious as that is, it is still easier to find footage of a bear running that has the angle/lighting conditions you need and composit it in to the footage of a skier than it is to CGI a believable bear. Realistic CGI hair/fur used to be extremely hard (both because we didn’t have the math for the simulations and because it takes a gigantic number of calculations) but that is pretty good for long hair and very good for short fur now. But it isn’t point-and-click either–you would have to spend a lot of hours tweaking dozens of settings in the hair (such as in Maya) and doing test renders to get it to look believable. Modeling a CGI bear from scratch takes a lot of skill, but fortunately premademodels already exist. Believable movement is also difficult (there is a reason that they put people in green suits covered with tennis balls to capture motion instead of just posing the models by hand.) If I, as a basement nerd but not a CGI artist, was attempting to CGI a bear to rotoscope into live video, I’d try creating the movements in Poser using one of those Poser models and export each frame of movement into a model that can be exported into Maya (because Poser’s renderer isn’t good enough for realism) and color/light each model in Maya and hope that the bear is distant/low resolution enough that the solid “hair” isn’t apparent. If you need more realistic appearance for the texture and movement of the hair, you would have to import the model “standing still” into your CGI program, skeleton it, weight the muscles, create the fur, and keyframe the movements, then do test renders and go back and redo things over and over until you get something acceptable. None of this can’t be done by a single individual with easy to get software (not necessarily cheap to get legally, but easy to get) and off-the-shelf PCs–the individual just has to take the time to learn the software and do the work. It might take hundreds of hours of work instead of the tens of hours for compositing, but if someone has plenty of free time and a passion for the project, it is something that can be done.
Removing an object from a video (as you mentioned) is the one thing that is pretty near point-and-click, as long as there is a good image of the background without the object and the background is fairly static. Fast changing background or static object with no “reference” background images? You are back to the range of “fairly hard” to “really hard” to fake.
tl;dr: Faking videos is not something that can’t be done by a knowledgeable, determined individual willing to do the tedious work, but it is also not something quick and easy.
When I started to play with editing - photo not video, I had to explain to people that you can’t just take an unwanted person out of a photo. I did do it for some of my own, but I had other pictures of the background to paste in; it was still tedious.
Green screen stuff can be fun. Get a video of a charging bear; film your ‘victim’ running in front of a green screen, and splice them. It comes out much the same as those old film scenes in a car, where the scenery is shown through the window. It would take hours to do though.
