Beyond skills: doing something different with voice assistants
Over the last 12 months, we have created some experiments with Voice Assistants, either driven by a specific brief or just by a technological need. For different reasons, we decided not to work on these projects beyond a rough prototype.
Nonetheless, we thought that each of them had something worthy of sharing, especially as they all illustrate interest ways to interact with voice assistants beyond the usual voice apps or skills. So here they are, three ideas on how to do something different with voice assistant.
1. Google Homie
What if your Google Home could also express through body language?
The way we interact with voice assistants is always through direct orders. Although their makers did their best to give them a realistic sounding voice, there is no space for etiquette when ordering to turn off the light or asking about the weather.
But if being rude to Amazon Alexa or Google Home is admissible, and actually less prone to misunderstandings then if adding any “please” or “thanks” to our commands, are we sure that getting used to being impolite to our voice assistant will not also make us impolite to each other?
We thought of an idea to make us more sensitive to how we communicate with voice assistants. We created a body for a Google Home device, and we made it react to good and bad words, by trembling with fear or swaying with joy. A subtle feature that doesn’t affect the interaction with the device, but an effective way to make us aware of our manners also when we communicate with voice assistants.
How did we do it
Google Homie is a Google Home assistant housed in a custom-made movable stand. To get the Google Home to move we didn’t interfere at all with the functioning of the original device but instead added another voice assistant, based on a Raspberry Pi and the Google AIY Voice Kit, that reacts with the same wake word of the Google Home.
Each time the user asks a command, the Raspberry Pi voice assistant checks for positive or negative words on the sentence and makes the Google Home sway or tremble by activating different motions for the servo motors on the stand.
What if you could be in control of when your Echo listens or not?
You could be forgiven for wondering if the AI-powered apocalypse was on its way, especially after Alexa’s unprompted, creepy laughter, and the timeAmazon Echo sent out a recording of a couple’s conversation,
That kind of behaviour certainly put a lot of people on edge.
In fact, the innocent age where we simply trusted products right out of the box is over. That’s why we thought of a way to give users direct control of their voice assistant.
Otto is a privacy-friendly add-on which sits on top of your Amazon Echo. It keeps the Echo microphone in mute, preventing it from listening until you want to use it.
By saying “Hey Otto” — or setting it up to respond to any other name — the device acts on the Echo buttons, freeing Alexa to attend to your every need. Once it’s responded, Otto puts the Echo back to the muted state until the next time.
How did we do it
The Echo has two physical buttons on its top, one to mute/unmute the device and another to activate the voice assistant without having to say “Alexa”. Otto consists of a speech recognition chip that can be trained to detect any word. Every time the chip recognises the word “Hey Otto”, it activates two solenoids that press the two buttons, to make the Echo ready to respond to commands instantly. Once the Echo has responded, the solenoids activate again to put the Echo back to mute.
3. The Dude (AKA the headphone trick)
What if your Echo could react to something different from voice?
One of the standard uses for a voice assistant is to let you control the home environment. On the contrary, there isn’t any simple way to get your Echo or Google Home react to what happens in the home. Any interaction with the device needs to start by somebody saying “Alexa” or “OK Google”, and then the request for the specific action.
But with hundreds of apps or skills for Echo and Google Home, it can be hard to remember the exact expression required for giving a command, or for some solutions, it feels just more interesting to initiate a conversation by an action rather than a voice command.
That’s why we came up with a simple trick to get an Amazon Echo to activate in reaction to a generic electronic signal rather than a voice command. This way you can, for instance, have Alexa greeting you when you come home, initiate the recipe skill when you open the cupboard or even, why not, talk to your dog when she’s looking for you in the bedroom.
Also, it’s a very simple way to give sophisticated voice control capabilities to any prototype.
How did we do it
The Dude trick involves a microcontroller solution that can store and reproduce audio files — a Bare Conductive Touchboard, for instance, is perfect for it — and a pair of headphones. You’ve probably guessed the idea by now, it simply consists of reproducing pre-recorded commands triggered by any kind of electronics. Then we created a custom skill for the Echo using Invocable, an online tool that allows you to prototype a voice app in no time through a visual programming language.
We’ve tested the trick on an Echo, and its microphones are so sensitive that we could keep the headphones volume to inaudible levels. Now the only thing left is to decide what sensor to use. In our experiment, we went simply with a motion sensor, activating a skill when a person is close to the echo, but the possibilities now are as vast as half of the Instructables website archive.