Pippytalk

A talking dog collar controlled by a Pebble watch.

Here's how it works:

Here's the collar in action:

Build it

This project has 5 components:

Hardware and collar: Follow Adafruit's amazing Phone Activated Talking Dog Collar project guide. Really, 95% of this project is following their instructions!
Code for the Adafruit Feather 32u4: see PippyCollar/PippyCollar.ino. This almost exactly matches the sketch in Adafruit's guide, but with an even simpler UART protocol.
iPhone app: see Pippytalk.xcodeproj. This is a hastily-modified version of nRF8001-Swift by Michael Teeuw. It forwards messages from the Pebble watch to the Feather board.
Pebble app: see pebble/. Use the Pebble SDK or CloudPebble to build and install. To install a pre-built version, click this link on your phone, and choose to open the file in the Pebble app.
Sounds: see sounds/. My samples are released under the MIT license; the two from "Up" are used without permission.

I hadn't done much soldering or circuit work before this. Maybe my mistakes will save you some headaches!

The output lines from the amp to the speaker must be a direct connection (as opposed to, say, attached by alligator clips). Something about high frequency signals requiring a good connection. Naturally this ended up being the last connection I chose to solder.
These grabber test clips are amazing for prototyping.
The "activity" pin on the Soundboard doesn’t work with super high bit rates. (The pin goes low when audio is playing, to sync the red LED to the sounds.) When I encoded my sounds with Ogg Vorbis at quality 10, the pin wouldn’t give feedback. I knew it must be the encoding since the LED worked with all of Adafruit’s samples. After I re-encoded my samples at a lower quality (6-8), the ACT pin worked correctly for all of them.
For two boards to communicate over UART, they must share a common ground, not be powered by two different USB batteries. If you think about voltage and signals for about 10 seconds, you'll understand why. Apparently I didn't have 10 seconds to spare during the few days I spent adding debug statements in Adafruit's Arduino libraries, cycling through boards, doubting my multimeter, and resoldering every joint.
If you are trying to sense "taps" (shakes) on the Pebble, know that vibrating the watch (vibe_short_pulse, etc.) can also trigger a tap. So if you vibrate the watch to give the user feedback when they shake their wrist, you may end up giving them a lot more feedback than intended.