CGI vs Practical - Can you tell the difference?

I view CGI to recreate this iconic practical effect from The Shining by the end of the video you'll discover which approach looks better which costs more and why CGI is so common today to start we import the photo into software called f spy which we use to dry out the axes guess the measurements and infer the focal length as this will just make it easier to model from then you bring that into your favorite 3D software package where you'll do most of the work in my case it's blender because it's free and open source and it's what my channel is about so hit subscribe and then watch my beginner series if you want to learn it now we just Trace what we see a hallway here a coffee table there maybe an over engineered oil heater over here sometimes you can buy models online but if you're recreating a set you probably need to make it from hand so that it's a precise match the easy stuff is usually simple hard surface objects and the difficult stuff is often organic and complex but all up a single lighter should be able to finish modeling everything in this scene in about two days next up lighting often this is done by capturing a reflective ball on set known as an hdri but if you don't have that which we don't we can make some guesses using accompanying reference photos of where the physical lights in the scene are then we need textures which if they weren't captured on location could be downloaded from a site like polygon which has common materials designed for 3D photorealism importantly each texture comes with these extra images known as Maps which are used by the 3D renderer to understand how the surface reacts to light setting up a material in 3D can be surprisingly time consuming but that's why smart artists use plugins so for example polygon provides a plugin which not only downloads and imports the texture but also sets up the material for you I might be biased but I use polygon all the time often the material is not an exact match so you use additional textures known as surface imperfections to layer on smudges scratches or whatever the photo reference says you need texturing often takes longer than modeling but it's still fairly predictable for a skilled artist texturing a scene like this shouldn't take longer than two to three days so we've got a pretty close replication but something is off all the main ingredients are there but this still looks like a 3D render the reason can be found plastered all around the Barnstorm VFX Studio grain Focus levels which according to them are the most common mistakes in VFX shots grain means adding noise to emulate the natural effect of film and digital camera sensors there's a common misconception that you can just render at low samples to create this grain but you can't at the very least you shouldn't it's splotchy contains fireflies it fluctuates across the frame depending on the object it's just ugly whereas camera sensor noise is typically uniform and more smooth across the frame so the correct way to emulate film grain is to First denoise all of that CG noise then use an After Effects plugin like film convert to add in film noise on top of it it sounds counter-intuitive but it makes a big difference Focus means matching the sharpness of the camera CG images are perfectly crisp and this is impossible in the real world so you enable depth of field set something as your point of focus then adjust the aperture till it matches the photo but even then the CG image will still look too sharp because no lens is perfect so a trick that I've just started using is to add a tiny amount of blur to the entire image again you wouldn't think this would do much but it's often the missing piece to realism and finally levels means using the waveform and histogram to ensure that there's the same amount of red green and Blues in each part of the frame now without a color calibrated reference to begin with we have one hand tie behind our back because it's impossible for us to know if the wall has a warm color to it or if it's the light or even the color grade so the best we can do is guess do lots of side-by-side comparisons looking at each object up close tweaking it and re-rendering it till the waveform matches as best we can so grain focus and levels before after our digital set is finally finished which means the real work now begins on the fluid simulation while Blended does have a fluid simulator it's so frustratingly unstable that almost nobody uses it thankfully though there's a paid add-on called flip fluid that works surprisingly well it's not as feature complete as the industry standard Houdini but also not as expensive every simulation starts the same way you set the boundary space tell it where the obstacles are and then you add an object to act as your fluid then you just push this magical bake button and wait about 10 minutes for the result problem with simulations though is that they never do quite what you'd expect so you fix all the problems wait another 10 minutes to see what you get then you identify all those new problems fix those try again and again and again tens or even hundreds of times you quickly learn that the aim of the game is to reduce the time it takes to simulate which you can do by making proxy low-res versions of every obstacle reducing the size of the boundary box and making sure fluid isn't getting trapped somewhere that it shouldn't and it's when the Finish Line seems so close that I typically spend the most amount of time you're confident that it's ready for the final bake so you increase the resolution to the highest setting you can stomach then wait days for the simulation to finish and then sigh as you spot one more mistake that requires a rebake but eventually after a solid six days of baking you end up with a simulation that doesn't have any glaring issues which is the best I can usually hope for so finally to make it look like blood we give it a gloss Shader and then add volume absorption so that the light terminates the deeper it penetrates for the final render the fluid got so dense at around frame 300 that my 430 90s ran out of GPU memory meaning I had to render the rest of the sequence on a much slower CPU so all up it was about four days of rendering at 15 000 samples of frame and to finish it just drag that sequence into Premiere add in the film grain and ladies and gentlemen we have our final shot [Music] foreign it to the original the winner is obviously practical I'm not the best 3D artist but arguably ilm is and even their recreation in Ready Player one doesn't look as good as the original which brings us back to our question if CG is usually inferior why is it so frequently chosen the first reason you can properly guess is cost according to a film director in La that I spoke with this shot would be about 50k on the very cheap end but more realistically about 100K plus due to unforeseen costs CG on the other hand is a fraction of the price According to some effects TDS I spoke with this could be done for about 14k on the cheap or 20K for feature quality or about one-fifth of the price and related to cost is risk because practical effects frequently go wrong a real house was set fire for a Russian Film called the sacrifice but right as they started filming the camera jammed meaning they got no footage and were forced to rebuild the entire house from scratch and shoot it again in Inception an explosive charge failed to detonate meaning this Tower fell the wrong way and ironically when they filmed the miniature version the exact same thing happened so in The Shining Stanley Kubrick was so anxious about this elevator scene that he wasn't even present for the shoot the entire investment hinged on every component working correctly as planned and luckily it did but it's no surprise then that today expensive action scenes are often done in CG where simulations can be rerun models can be reused or sets redesigned on the dollar another Advantage is flexibility if Stanley made the crushing realization that the camera was actually too far from the action he would have no choice but to reshoot it see G on the other hand is flexible to a fault you want the camera to move no problem you want the lights to flicker easy it's so flexible it's actually led to a game that fill tip it calls find what's wrong with this shot but misuse aside flexibility is a big reason why CG has chosen to begin with the final Advantage is safety in the movie you'll notice that while we see Shelley then the blood then Shelley we never see both in the same shot which actually creates a problem for the audience because we don't know if this is real or if she hallucinated it which is why I think this shot could be improved if it was CG because you could have Shelley run right up to the elevator the doors open you switch to a digital double and we see her get washed away just like in Ready Player one it's rarely talked about but safety is one of the most valuable uses for CG today padding is removed sets a heightened and digital humans are stopped essentially letting us have our cake and eat it too feels dangerous and nobody died but for all the advantages to CG there is one major disadvantage thrill part of the reason Jackie Chan films are so fun to watch is because you know it's real There's real pain real danger and real skill today's audiences don't see a lot of that they know that if there's a slight risk of an accident or it's too difficult or it's cheaper then it'll probably just be shot on green screen so when I watched this scene from extraction 2 I actually didn't think much of it a clever trick probably done with a crane a shaky platform and a green screen until I watch some promotional material and discover that it was real they really just landed a helicopter on a moving train now I look at the shot differently the result hasn't changed but my knowledge of how it was filmed has changed my enjoyment this isn't acting everybody here is really scared and it took extreme skill by the pilot to pull it off in essence I can relate to what I'm seeing she Studios have realized this that if an audience knows a stunt was shot practically it not only makes the shot more interesting but it will probably sell more tickets which is why we're now told through copious amounts of marketing what was shot practically even if it's only half true I mean what's the coolest story I simulated this on my computer or they really flooded a hotel with blood which is why it might not even matter that CG is cheaper more flexible and safer because of the belief that it's practical gets more Buds and seats then it might be worth doing for real they just need to make sure that they promote that one stump that was practical and not talk about the rest of the movie that's why I think CG is a lot like a magic trick the audience doesn't actually want to know how it's done and provided you don't tell them they'll happily believe that they just witnessed something incredible which in the movie industry is what it's all about if you enjoyed this video you might like this one where I remade an abandoned house from The Last of Us click here to watch it and I'll see you next time

Channel: Blender Guru

Views: 5,364,345

Keywords: blender, tutorial, andrew price, blender guru, blender tutorial, the shining, elevator, the shining elevator scene, practical vs cgi, practical fx

Id: mP-idHLydNQ

Length: 11min 21sec (681 seconds)

Published: Sun Sep 17 2023