The better CGI gets, the worse it gets

This Cracked article is devoted to taking an axe to the trailers for Jurassic World and Terminator: Genesys, detailing why everyone is saying they look fake as hell. Author David Christopher Bell further goes into why their forebears (Jurassic Park, T2) from the primordial days of CGI in the early ’90s looked so much better.

As further evidence, here is an article from an industry website read by digital artists themselves. In their top-5 list of all-time greatest CGI movies, the only one produced in the last 10 years was Avatar. Their #1 choice was, yep, the original Jurassic Park, which is old enough to drink. And this is a website by and for industry professionals. Even they acknowledge their current work has yet to beat a movie that came out in an era where graphics workstations had a fraction of the CPU power of a freaking iPhone.

It’s a shame the Cracked article made no mention of Star Wars, because no article about CGI failure is complete without mentioning George Lucas and his execrable prequels.

A comprehensive list of why, exactly, the prequels suck would make for a book that looks like this:

… but let’s start with George’s insane overuse of CGI. During press interviews of the time, he could not talk about anything but CGI. He went on and on about how glad he had CGI when he was forced to use models and on-scene live shooting for the originals.

But as sophisticated as computer imaging has gotten today, let alone the turn-of-the-millenium era of the prequels, it’s still hard to fool the human brain into thinking that it’s looking at something real. For instance, insane amounts of talent, computer hardware and man-hours went into the creation of Gollum for the Lord of the Rings movies. It certainly was a technical marvel… but my brain did not buy for one second that Frodo and Sam were interacting with anything but a cartoon. It’s not that I’m somehow incapable of suspension of disbelief — I was totally buying that Helm’s Deep was under siege by 10,000 orcs, for instance. But then again, the orcs we see up close were played by real actors with a crapload of makeup. I even knew the actor who played the one orc who ran up with that torch to light the bombs.

(At parties, he loves hearing: “Anyone got a light?”)

But do you think you or I would have bought into that army of orcs if they were all CGI and the whole movie was shot on green-screen? Which is what would have happened if Lucas was running the show? Hell no.

Going back to Star Wars: What Lucas never realized is that despite how low-tech the originals were, they were far more convincing than his later monstrosities, and not just because of the prequel’s abdominal writing or wooden acting or casual racism. Forget the characters — let’s start with just the locations. Tatooine (shot in real-life Tunisia) and Hoth (Norway) looked far more convincing than whatever-the-hell planet Natalie Portman’s character is from. Or take Coruscant from the second prequel, which looked like the background to a video game and could not match Blade Runner (1982) in establishing a futuristic megalopolis. Or how about space? It’s called “Star Wars,” after all. And giant space battles should be where CGI shines. Right? The following two pics are screengrabs from Youtube clips of Return of the Jedi (1983) and Revenge of the Sith (2005):

Never mind the vast gulf of quality in plotting and writing between the two scenes, except let it suffice to say that most of the “plot” in the latter sequence involves trying to get these cutsey spider-robot things off of Ewan McGregor’s fighter craft. Ugh. But we’re just here to discuss the special effects, which were the whole point of the prequels according to Lucas himself.

On the top screenshot, you can clearly tell what’s going on and with even minimal knowledge of the films, can tell which ship belongs to which side and which way they’re moving. Notice how brilliantly the X-wing’s red engines stand out. Each ship comes out crisp and clean, and our brains have no problem thinking of them as “real.”

On the bottom shot… wtf? Besides the weird, gauzy colorization so characteristic of CGI-abusing films (space is black, Lucas, not purple), each ship seems hazy and indistinct. Besides the heroes’ two fighters and the recognizable ship on the upper left, good luck making heads or tails of any of the others. And the screenshot may as well be from a video game — our brains do not even entertain the illusion that this is “real.”

Same with Jar Jar Binks. The character was ridiculous on many levels, but one was seeing Liam Neeson trying to emote to a goddamn cartoon character. Was Wakko Warner not available?

In any event, the word is that Jurassic World is to Jurassic Park, what the Star Wars prequels were to the original trilogy. That’s sad. But that’s also 12 bucks they are saving me.

As for Episode 7? Too early to tell for sure, of course. And I am anything but a fan of JJ Abrams. But they claim they limited the CGI, and R2-D2 certainly looks more real than he did in the prequels, which is something. And the dark stormtrooper with the cloak in the trailer is obviously a real actor in a costume, as opposed to Lucas’ hordes of CGI robot and clone troopers. But the white-armored Stormtrooper army we get a glimpse of looks worryingly fake — let’s hope it is just an establishing shot.

The brain is pretty good at calling bullshit on CGI, far more so than it is on models. Lucas never realized that being forced to use models in the ’70s and ’80s represented a big reason why the franchise was a success, as if he had had CGI back then to abuse, his movies would have been laughed out of the theaters. Sadly, many movie-makers today have still refused to learn from his failures.

But at least the lesson took on the small screen: CGI is limited even on a lavish show like Game of Thrones, and is used to enhance backgrounds of real shots more than anything else not involving dragons. The blood-and-bones of the series’ production involves old-fashioned location shoots and famously obsessive costuming and makeup. And I think fans are pretty OK with the results.