Last week the fourth episode of Game of Thrones' seventh series leaked online, thanks to the fact an Indian broadcaster managed to lose the tape. Plenty of people are desperate for more of George RR Martin's fantasy world, so no doubt many of them will have watched the leak - regardless of the fact that it wasn't the best resolution in the world.
Some people who watched the leak might have noticed that the runtime seemed rather short, given how there was only 47 minutes of programming in the leaked file - that'd including the opening and closing credits. The actual episode broadcast on HBO (and simulcast around the world) last night was 49 minutes long. So what gives?
The first assumption might be that some scenes were cut, possibly those with any nudity, which isn't an uncommon practise in India. But the answer is a lot less frustrating than that. The leaked episode is just running faster, and you can hear it in the characters' voices.
We have compared both versions of the episode side-by-side, specifically the part where Bronn and Jaime are having a horseback conversation (pictured above). In an ideal world we would show you the two together, but that could land us in some hot water with HBO's copyright people. So we haven't.
We cut out 23 seconds of the conversation from the HBO-broadcast version of the episode, which ended up being 1-2 seconds longer than the same piece of conversation cut from the leak. Proving that the leak was running ever-so-slightly faster. Their voices were noticeably higher, too. Not by much, but enough that you could hear something was off.
But why would this happen? A few reasons, as it turns out.
In recent years broadcasters, particularly those in the US, have been caught speeding up syndicated TV programmes so they have space to run more adverts during the breaks. To the point where some episodes of Seinfeld drop from a 25-minute runtime to 22-minutes.
This also has happened in the UK, with American-made film and TV programmes running slightly faster on British screens. This is less to do with ad revenue and more to do with the differences between NTSC and PAL. PAL TV standards play everything at 25 FPS, whereas most stuff is actually shot in 24FPS. This means that PAL video has to be run slightly faster, by a rate of around 4.16%. This ends up with the video in question being four per cent shorter, and has a higher pitch. Meanwhile NTSC televisions play back at 30FPS, which is the opposite. Every other frame of the 24FPS footage ends up being shown for slightly longer (at a ratio of 3:2) That means everything plays for longer, and there's no problem of raised pitch.
While the differentiation between PAL and NTSC has lessened over the years, particularly with the widespread adoption of digital broadcasting, they apparently do still matter. India does use PAL video, as it turns out, and my maths works out that four per cent of 49 is 1.96, which would explain the two minute discrepancy between the two versions of this episode.