Less Is More – Codemanship's Blog

Skip to content

Jurassic Park (1993) has about 60 visual effects shots containing computer-generated elements.

In 1993, photorealistic, cinema-quality CGI video took much longer and cost much more to produce than it would today. The high-end Silicon Graphics computers used by Industrial Light & Magic took 10-12 hours to render each frame. A lower-end smartphone could do it in 2-3 seconds today.

As the technology got faster – and the software got better and easier to use (think of what kids can do on laptops today) – the cost of CGI shots plummeted.

This was great news, because now studios needed far fewer CGI artists and technicians to make their movies. And that’s totally what happened. All the visual effects for Avengers: End Game were done by one guy called Barry working Tuesdays part-time.

Jurassic World has 1,000 computer-generated shots – 16x as many – and each frame contains about 1,000x as many polygons being animated and rendered.

But, bafflingly, Jurassic World only has a 6.9 rating on IMDB, compared to Jurassic Park’s 8.2 rating – even though Jurassic World is obviously 16,000x as good.

Could it be, perhaps, that more CGI with more polygons produced by more computing power isn’t necessarily what excites audiences? I remember every CGI shot in Jurassic Park. I struggle to remember any specific shots in Jurassic World, CGI or otherwise.

Maybe the creators of that sequel were so preoccupied with whether or not they could do more CGI shots with more elements and more detail, they didn’t stop to think if they should.

During my career, the computing power available has similarly exploded. The laptop I use day-to-day is easily 10,000x as powerful as the 486 I started on, and the tools I use are far more advanced. Compiles that used to take 15-20 minutes now finish in 15-20 seconds. I can run 10,000 unit tests in 2-3 minutes, whereas in the 90s I’d run a suite that big overnight.

But, like with CGI in movies, this has not produced a proportional increase in productivity or in value created. It could be argued that the very low cost of computation today has made little difference at all. Sure, there are way more developers now, creating way more software. But is it software anybody really wants or needs? Can society absorb that much software that fast?

The high cost of CGI in Jurassic Park meant that the filmmakers had to make every shot count. Just a couple of years earlier, the makers of Terminator 2: Judgement Day (IMDB rating: 8.6) had to make even more with even less.

Recently-published research found that the more elements in a shot, the less viewers paid attention. The brain only has so much visual bandwidth, and modern CGI-heavy action movies tend to overload viewers with visual information – making the overall experience more "Meh" than "Wow" (thanks to Rob Bowley for the inspiration).

And FX budgets haven’t shrunk with the cost of compute, either. They’ve ballooned. The average Marvel fare now has the equivalent of a whole town working on dozens of elements in thousands of shots. Jevons paradox in action.

So all this extra compute was for naught, Jason? Not quite.

There are some shining examples where the power was used not to create More StuffTM, but to create better stuff, and to create it cheaper.

Monsters (2010) cost famously little to produce, and the 200-or-so CGI effects shots really were done by one guy (called Gareth, working full-time) on consumer hardware. The whole movie cost about £300,000 to make. Here’s the thing, though – Gareth Edwards knew what he was doing .

He knew how to use the technology, having busked as a VFX artist in television. But, more importantly, he knew about directing, and about cinematography, and about editing, and about story. It might not be the greatest sci-fi movie of all time, but for the price of a semi-detached house, it’s pretty miraculous.

James Cameron is another director who started his career doing a heck of a lot with not very much for famously frugal producer Roger Corman, and is equally well-versed in key technical disciplines as well as the creative ones. Ever wondered why the Avatar films – being almost entirely CGI – look so much better than most modern FX-packed movies?

The evidence seems to suggest that the most valuable way we can leverage all this extra computing power is not necessarily to create more, but to iterate and refine faster to bigger impacts. What’s come down – should we choose to use it – is the cost of getting it wrong and trying again (and again).

Compiling and running tests faster doesn’t mean I can ship more features. It means I can iterate features more times and learn more from that for the same cost.

And what matters more than ever are human skill, judgement and taste. That will always be in short supply.

Otherwise, we’re just shipping Jurassic Worlds faster than the audience can absorb them.

Wanna see the receipt?

Share this:

Share on X (Opens in new...