Shake it off: Remembering Apple’s Academy Award–winning failure
People are reminiscing about 25 years of the iMac, which is fine and all, but I’m doing to do something a little different. I’m reminiscing about 21 years of Apple’s Shake.
What’s that? You have no idea what I’m talking about? You just want candy-colored computers? Too bad. You’re going to learn something you’ll never need to know, whether you like it or not.
In February of 2002 Apple acquired Nothing Real, a software company based in Venice, California that made the industry-leading Shake compositing software. It was in that weird period where Apple was picking up steam, and trying to convert PC users to Mac users. There was no Shake version for the Mac, but there was for Linux, Windows, and Irix (RIP).
Just a few months after that acquisition, Apple released Shake 2.5 for the Mac (along with the existing supported platforms). Shake 2.5 would be the final version for Windows. Apple’s goal was pretty clear just from the press release—Shake cost half as much on Mac as it would on another platforms. (And this is a product that cost $10,000 per seat.) By Grabthar’s hammer, what a savings!
Apple was trying to convert high-end professionals, who would theoretically buy high-end Macs. As with Apple’s acquisitions of Final Cut Pro and Logic Pro, here was another pool of potential Mac buyers. (For context, this is the same year Apple released the second-generation iPod that added Windows support in an effort to get people to use Apple products in their personal lives.)
There were some problems with Apple’s strategy, though. While Apple managed to kill Shake for Windows, it could never do the same to the Linux version. That’s because all the big VFX houses were doing their work using Linux. It was easier to convert an editor or a small group of editors to Final Cut Pro than it was to pitch a corporation on converting hundreds of desks, and their render farms, to Power Macs or Mac Pros—even with the steep software discounts. (Not even Pixar!)
For Mac users, Shake never made any effort to fit in. It was born of an era where professional software tended to have its own custom interfaces so that the apps could run on a variety of platforms and work more or less the same. It never even adopted the Mac’s file open and save dialogs, buttons, or anything. It had bevels out the wazoo, which was the style at the time. It had some 3D-effect turquoise slider elements, but I wouldn’t call the interface “lickable.” Don’t lick the bevels, kids.
This custom UI was part of the reason it could be quickly ported to Apple’s Mac OS X Unix-compatible system from Linux and Irix, because no one needed to worry about how interface elements would be placed. It was literally the same interface.
Nodes and noodles
While Shake was not the first piece of software with a node-graph interface, it was arguably the most widely available at the time of its initial release.1
The benefits of a node-based approach might not be clear at first glance. After all, when most people think of image editing, they think about it linearly: You open a file, you change some colors, add a blur, and save your file. (Or, alternately, consider a bunch of vertically stacked Photoshop layers.)
When it comes to video compositing, things get a lot more complicated. You might need to re-use the same element twice, or combine many elements instead of just editing one. And what if all your work needs to be done over a thousand frames of moving footage? A linear workflow just won’t do. You need to have the ability to branch and merge your work.
The node graph allows for the visual representation of complex, interconnected assets and adjustments. It allows for multiple inputs and outputs. It allows for reusing work, where the file being operated on can be swapped out for some other file while still applying the same edits.
This interface also helps you evaluate the output at any point along the line, including walking up or down the tree to see which node is causing a certain effect. This comes in handy when you’re working on version 100 and your client has decided that the thing you added 80 versions ago is something they want to take out. You can’t just keep pressing Command-Z—you need to alter just that one choice.
This was real non-destructive editing. And it was powerful.
It’s also important to note that this interface wasn’t a stack (like Apple’s Shortcuts, or Photoshop’s History palette) that runs from top to bottom. It was a web of nodes, all connected to one another. Shake visualized these connections between nodes with noodles—not straight lines, but a curving piece of spaghetti. (This is whimsy, as manifested by the developers of high-end professional software.) If you preferred straight lines, and you could straighten out those noodles by adjusting a preference called Noodle Tension. (I’m not making that up.) As for me, I’m a monster. I love my curvy noodles.
Everything flowed logically. Nodes that generated images didn’t have a little input bump on the top of the node, because they didn’t accept input, only generated output. Nodes that did operations had at least one input and one output. Any time there was more than one input or output, there would be a discreet and separate connection point for each so that things didn’t all collide in one spot. Each of these nodes also had a mask input, where any kind of alpha could be connected to limit the area that the node was editing. It was sort of like a Clipping Mask in Photoshop, but way easier to reuse and adjust for multiple nodes at once.
Shake even had a pretty comprehensive set of paint tools, where paint strokes were all saved nondestructively. There were warping tools that were used to create talking animals. All those things were there, editable, non-destructive, and portable.
And if you ever got a collection of nodes that you used a lot, you could group them and save them as a macro, which would appear as a single node. This was a great way to save time and reduce visual complexity.
The namesake feature of Shake was that if you wanted to remove a node and take it somewhere else, instead of disconnecting every noodle going in and out of it, and reconnecting those up and downstream, you could just hold down on the mouse button, and shake the node around until it until it was released. It wasn’t something you needed to do a lot of the time, but when you did do it, it was charming (and you were very relieved you didn’t need to reconnect things).
Multiple endings for Shake
From Apple’s March 2004 press release on Lord of the Rings: The Return of the King winning the Academy Award for Best Visual Effects in a Motion Picture:
“We’re thrilled that for seven years in a row, movies created with Shake have won the Oscar for best visual effects,” said Steve Jobs, Apple’s CEO. “Shake is helping Hollywood film editors communicate their vision and deliver their art at an Academy Award winning level. We couldn’t be happier.”
It’s always been weird to me that Steve Jobs referred to them as editors, but maybe that’s because of how Shake was often positioned as a companion to Final Cut. (It didn’t work anything like Final Cut, but whatever!) If you were a small shop that had Final Cut Pro and needed to do a few effects, than it was great to have this class-leading software available to you, even if you weren’t making a “Lord of the Rings” movie.
But Shake just didn’t pan out the same way Apple’s purchases of Final Cut Pro and Logic Pro did. While it was wildly popular, award-winning software, it was only ever wildly popular and award-winning on other platforms. The last major version of Shake was released in 2006, with a minor update was released in 2008. Shake’s price was slashed and slashed again over its lifetime, with the final offer being a $50,000 site license so a VFX house could intall it on as many Linux boxes as they could muster. (That would have been five Linux workstation licenses under the original Shake pricing.)
This meant that I got to use Shake once again at Imageworks on Watchmen, including a wild day and night where I was working in a Shake script on Dr. Manhattan, and then copying and saving those changes to the main comp script where the plate photography was being stitched together for a bunch of mobsters exploding. The other compositor would tell me when he was done, and I’d copy more nodes and adjustments to the file. Back and forth. The workflow was still powerful, even if it hadn’t seen active development in three years.
Shake was unceremoniously discontinued in 2009. Apple wasn’t the company trying to convert visual effects houses to buy cheese graters any longer. It had become a company printing money from selling iPhones. Shake’s developers were moved on to other projects, or departed the company altogether.
I know there was never a financial case for Apple to have continued development of Shake. But it’s sad to see that misguided acquisition kill a product that was formative for me, and others. (Christa Mrgan mentioned it as an influence on Audio Hijack 3’s redesign.
Fortunately, node-based compositing didn’t die with Shake—far from it. Foundry helped “productize” Digital Domain’s Nuke just as Shake died, and that’s now the industry standard.
Apple was the wrong company to acquire Shake and the wrong company to attempt to be an industry leader in tools for visual effects artists. Sadly, Shake’s legacy isn’t about those smart nodes and fun noodles—it’s as a cautionary tale about misguided acquisitions.
- The original Shake design sketches have been uploaded to Flickr by Nothing Real co-founder Ron Brinkmann. ↩
[Joe Rosensteel is a VFX artist and writer based in Los Angeles.]