If you write a good book, in a hundred years, people will still be able to read and appreciate it. If you create an app, however successful it is, in a few years it will be obsolete. Periscope was a ground-breaking live video app when it launched in 2015, but it was shut down some years ago now and, like most obsolete software, is now largely forgotten.
My involvement with Periscope began in early 2013. I had been working as a freelancer for some decades, specialising in video software for Windows, but over time, demand for Windows expertise declined as the iPhone took off. Finding myself with a few quiet weeks with little consulting work, I spent some time looking at video software on the iPhone.
At the time, the iPhone did not support low-latency video streaming: the video encoder could only write to a file, rather than sending data directly to a video streaming output, meaning that the video would be delayed by at least 10 seconds. It occurred to me that, if you understood the file format, you could extract the video from the file while it was still being written to the file, and this would allow you to have much lower-latency video.
I wrote a sample app to demonstrate this, and published the source on my website, confident that this demonstration of expertise would bring in plenty of enquiries about iPhone software, but there was nothing. I went back to writing Windows software and didn’t think any more about it.
About a year later, in early 2014, two young Stanford graduates, Joe Bernstein and Kayvon Beykpour, got in touch with me. They had been working on a photo-sharing app, but were not happy with it and had decided to pivot to video. They wanted to be able to broadcast live from an iPhone using the app, and for viewers to able to watch the live broadcast, and respond with comments and hearts. However, the live feedback from the viewers would only work if they were seeing the video within a few seconds of the broadcast — watching video with a 10+ second delay would not work. So Joe and Kayvon needed a way to stream low-latency live video on the iPhone, but weren’t quite sure how to go about it. I knew how to do it, but had not had a use for it. Joe and Kayvon asked me to spend a few days helping them get started, using my sample code as a starting point. Those few days ended up being eight years of working together.
Prior to Periscope, there were two forms of video streaming. Video conferencing, with a two-way conversation, required extremely short delays — the delay from camera to viewer needs to be in the tens of milliseconds, and this imposes significant restrictions on the video. Other forms of streaming were one-way and there was no need for low latency. Like TV today, the delays from camera to viewer can be 60 seconds or more. Periscope’s hearts and comments feedback mechanism required a delay of around 2-5 seconds. There was no existing model for this, so we had to experiment.
Downloading data from the internet on a mobile phone can be quite slow, but uploading to the internet is much slower — typically five times slower than downloading. The upload bandwidth is usually just about sufficient to stream good quality audio, but a long way short of the bandwidth that video requires. Also mobile phones are mobile, and when moving, the bandwidth varies and the connection comes and goes. Most of my time over the first few months was spent working on ways to estimate the available bandwidth and ways to ensure we got the best video quality we could.
This was made harder by my location in rural Wales. At our house, we had no mobile phone signal at all. But if I walked ten minutes along a footpath, there was a spot in the middle of fields where the land fell away and there was a signal from a mobile phone mast in the distance. In the summer of 2014, I spent many hours in that field, accompanied by our spaniel, Jack. I sat on a wall with my laptop and two mobile phones, broadcasting from one to the other.
By the New Year, we had an app that worked well, with dozens of beta testers broadcasting live every day, and we were ready to start thinking about launch. But the acquisition by Twitter changed our perspective. The attention and visibility would bring a lot of users to Periscope on launch day, and we spent several weeks making sure we were ready for the influx. We expected that a small number of well-known people would go live, and far more people would watch. When we did launch, we realised that we had got it wrong. Most people wanted to broadcast, not to watch, and the app was completely swamped by the number of simultaneous broadcasts. My colleague Aaron had worked on a nicely animated rotating globe that would be shown when there were no live broadcasts, but it never appeared.
For a while, Periscope was extremely popular. You could watch broadcasts from all around the globe — I once followed the sunset from one broadcast to another around the world. For some people, for example cancer patients stuck in hospital, it provided a connection to the outside world. And it was at its best in breaking news — even the BBC used it from time to time. In December 2015, it was Apple’s App Of The Year, and was number one in the Guardian’s list of the best apps on iPhone and Android. In 2021, it was shut down, as Twitter’s fragility forced managers to concentrate on a single app and a single user base.