How we built a tweeting robot that took over Twitter to the tune of 3100 tweets an hour at its maximum rate.
The group here at Paper Leaf has a few things in common: a love for design; a penchant to use (overuse?) social media; and the general “tinkerer/hacker” personality type. What happens, then, when you add them all up and mix in an internal Build Day?
You get a tweeting robot. Meet Hugo:
If you tweet with the hashtag #hugorobot, when Hugo is awake, Hugo will read it out loud. So, for example, if you tweeted “I don’t know, that logo needs to be bigger I think #hugorobot”, Hugo will say, out loud to everyone, “I don’t know, that logo needs to be bigger I think”.
That’s the gist. We launched him on December 2, 2015, and it quickly took over Twitter to the tune of 3100 tweets an hour at its maximum rate. We livestreamed the whole thing, and the edited livestream video and more launch campaign data is below.
Um, more like why not a tweeting robot.
As the video above explains, Hugo came about in a brainstorming session for one of our internal Build Days here at Paper Leaf. In those Build Days, we aim to work on something fun and challenging together; something that might push our boundaries a bit. In this instance, we knew we wanted to work on something that mixed hardware and social media (hat tip to ISL, who kinda pioneered this stuff). In the end, we came up with the idea of a tweeting robot that reads tweets out loud… because, fun.
Goal-wise, we wanted Hugo to be a self-contained unit that could be moved around easily. Ideally he’d only need a WiFi connection, and could be controlled via tweets. That way, we could lend him out to other friends if they wanted to have Hugo for a day (or campaign).
Hugo runs off of a little battery pack, which is connected to a Raspberry Pi. The Pi has a WiFi dongle attached, to connect to a network, and is running a custom Python script that monitors Twitter for a specific hashtag – in this case, #hugorobot. The script grabs those tweets, strips out some stuff (e.g. URLs and hashtags), then runs the tweet through a text-to-speech engine. That, in turn, gets broadcast out of a little speaker connected to the Pi that is a part of Hugo’s chest.
Pretty nifty stuff.
Physically, Hugo was designed in Illustrator to accommodate all his pieces. He’s constructed out of foamcore, which was then puttied, sanded, and painted – because hey, even robots need to look good. In the future, we might end up 3D-printing him or building him out of wood & steel for his nights on the town. But for version 1.0, foamcore does the trick.
We also added in some extra controls. Hugo doesn’t read racist tweets or a variety of other not-cool words / phrases – but he does swear – and can in fact be further programmed to avoid any variety of tweets that have blacklisted words in them. Hugo can also be controlled via tweets from his own Twitter account via a variety of hashtag-based commands. For example:
Note that these have to come from Hugo’s account for them to work, so sorry – you can’t shut down Hugo. We have a few others in there, for debugging, but those are some of the more interesting ones.
Like we mentioned above, Hugo launched on the morning of December 2, 2015 alongside a livestream (so people could see and hear the results of their tweets) and quickly went viral. Here’s an edited version of the livestream video, showing Hugo reading the best tweets of the day out loud to our office:
The whole campaign took the design & marketing industries by storm – Fast Company wrote a post about the campaign, which AIGA shared along with hundreds and hundreds of others.
Here are some interesting tidbits we learned in the 4.5 hours the #hugorobot experiment was running:
Overall, Hugo was a huge success and an interesting project from start to finish. But he’s not really finished… I’m sure we’ll see more of him here and there.
Think you have an event or campaign that could use a mashup of social media and hardware? Drop us a line.