This cute, friendly creature looks like a bird, sings bird songs, and behaves like a bird, incorporating Perlin Noise into its darting blackbird-inspired motions. Taking the form of a golden oriole, Birdveillance awakens from its slumber to follow your face with a playful coquettishness as you walk by. You look up for a moment to determine the source of this indoor birdcall, and might even chat with the bird itself, before resuming your human activities in the hallway. But at that moment, a light flashes! The bird tweets! And the Twitter account @birdveillance Tweets a photograph of your face, accompanied by a speech-to-text snippet of your conversation.







Once the twitter handle is revealed (via a nametag on the bird's nest), what once seemed like an innocent bird takes on a more complex character as the user observes how it publicly tweets moments that should have remained private.







Once people understand that they were being watched/recorded, they may approach the bird differently, attempting to control the bird's perception. Instead of being a passive, oblivious surveillance subject, the user now takes an active role, either averting eyes with the bird, or guiding the conversation until they see themselves portrayed publicly in the way they want to be portrayed.

Background
"Fundamentally, surveillance is a business model of the Internet," according to Bruce Schneier, a cryptographer who is attempting to rally internet engineers to fight back. We thought that a bird who "tweets" in both senses of the word would be a great way to draw attention to surveillance and make us all more aware of the ways that formerly private information is increasingly accessible online, whether it's via NSA spying or tweets of our own volition.




We played with the surveillance cameras that ITP has, and almost worked with one of those before deciding it was way too heavy and bulky. In appearance, we wanted to emphasize bird over surveillance camera. We did a lot of user testing and found that the way the bird looks would be the key factor, along with the sound, to determine the user's initial approach towards what is essentially a camera and a couple motors. So we worked hard to hide the mechanics and technology and researched how birds look, sound and move to try to recreate this in a stripped-down, simple, fun and playful way.

Audience
We imagined this as an installation on the floor of ITP. But it could be reinstalled in a variety of spaces where there are many people passing by.

User Scenario
Described above in Description, but in addition to the sound and flash, we might set up a little bird feeder and invite people to "Feed the Bird" by putting bird feed into a feeder so that they look at the bird to get the bird's attention.

Implementation
Rather than utilize an existing stuffed bird, we decided to build our own. The head is a plastic apple with a webcam poking out of the beak-hole, LED eyes, a bunch of cables sneaking out the back, and screws connecting it to a Pan/Tilt bracket. The pan/tilt has two servo motors connected to an Arduino. The bracket is affixed to a neck/spine made of a piece of a wooden bed frame that Jason found in somebody's trash. The spine sits in a plastic cookie jar. The cookie jar has a speaker in its belly. The base is weighed down with a flour-clay mixture plus a broken power supply from the scrap shelf to keep the bird's butt from sliding around. The speaker and webcam are connected to a laptop running Processing that is hidden beneath the bird's nest. The bird's cuts are covered in feathers and toe socks. The bird would ideally sit in a nest or a sturdy branch that we'd still need to build, and the nest would have a laser-cut name plate: "@birdveillance" so people know to check out its twitter. We're also looking into a "flash" in addition to the LED's so that it's clear when the bird snaps a photo, and this flash might also draw attention from across the room. We need to tweak our microphone setup if we're going to do quality speech-to-text, but evne if the bird just tweets random nonsequitors from an array of strings it overheard over the course of the day, that's fun too.

Conclusion
We learned a lot about programming, interaction, and especially fabrication. We broke a servo motor that didn't have enough torque before figuring out how to fasten a larger more powerful one into the tiny pan/tilt bracket. We learned that twitter has a daily photo upload limit so if we do this for the show we'll have to upload the photos to our own server and link them in the tweets. We also learned that although it's possible to fit the sounds we want to play onto an Arduino's 32kb flash memory using Huffman encoding and the Mozzi library, it's not possible to run this and the Servo motor library at the same time cuz both use timers. So we're running sound from the same Processing sketch as the OpenCV face tracking.