If you're one of the tens of millions who has watched the band OK Go's elaborate music videos, you've probably thought two things: "Wow!" and, "How did they do that?"
From dancing on treadmills, to performing in zero gravity on a Russian plane, the LA band has become as well known for its ambitious one-take videos as the songs they accompany.
"We want to address this question we get asked all the time, but we've really never come up with an adequate answer for," said singer Damian Kulash on stage at the TED Conference in Vancouver this week.
"How do we come up with those ideas?"
What they don't do is sit just around and think, said Kulash in an interview.
"If you plan in advance, then you're limited to the ideas that you can have in advance, which means that you're going to have the same boring ideas that everyone else has."
The problem, said Kulash, is finding ideas that are both predictable to execute and surprising to observe.
Imagine 130 moving parts in a chain reaction.
Getting each step to be 90 per cent reliable might sound like good odds, but "it's actually terrible," said Kulash, who showed the math on stage.
"The chance of getting all 130 things to not fail at the same time ... is literally one in a million."
So, he explained, they need ideas that work 100 per cent of the time, which normally means they've been done before.
"Unless you put yourself in a position where you can discover things ... that you didn't know were reliable before you started."
Step one: play
The solution, he says, is to play first.
In the Grammy-winning video for Here it Goes Again, where they dance on treadmills, the band and Kulash's sister, film director Trish Sie, "hung out with eight treadmills for two weeks" before planning the video, he said.
"We just played on them."
As a result, they discovered Kulash's favourite move, where they looked like they were ice skating as they moved from treadmill to treadmill.
"There's no way I could have predicted this ice skating move," which was easy to repeat, he said.
"It's the kind of thing your brain can't work out until you just try it."
A week of play at zero gravity
Sounds easy with treadmills on Earth, but what about on a plane that's plummeting through the sky to give the sensation of zero gravity?
They played there too, said Kulash, spending one-third of their budget on six flights and a week of experimentation.
"There's so many weird little lessons you learn," said Kulash.
Everyone tries to swim in zero gravity, so they had to learn not to.
Then came trial and error. Did a string of beads or a chain move in a more interesting way?
What about a house of playing cards? That didn't work, as air currents kept knocking the light cards around.
Planning starts with different ideas
After a week of play, they took a month to watch their footage and figure out what popped and would deliver the "wow" while being repeatable.
Some things looked cooler on video than they had seemed at the time.
Others, the opposite: "That felt so cool, but it just looks dumb, it looks exactly like it does on Earth."
They chose their favourite 10 moves, then others to fill in the gaps and make the arc of the video's choreography.
He said the process is natural for musicians, who wouldn't plan their notes by the microsecond without playing them first.
But in the normal film world it can be tough to swallow, he says.
"The planning basically starts once you have a different set of ideas than you would have to start with," he said.
"It can cause difficulty ... but a lot of times you get much more interesting results."