Alexa, let's play a game!

"Alexa, let's play a game!"

How we came up with the idea to develop Campfire, a platform for interactive audio games for smart speakers. By Torben Ludwig, CEO and co-founder of Frozen Donkey.

When Amazon launched Alexa, the first mass-market smart speaker, I didn't think "Oh, I'll be great at ordering printer paper with this", but "Wow, this opens up a whole new way to tell interactive stories".
The spark needed to turn this idea into a reality manifested itself during a visit to Disneyland Paris, where I rode through miniaturised landscapes and watched fairy tales live. Even while waiting in line, it occurred to me that it would be amazing to be able to personally interact with professionally spoken stories and thus make them more tangible.

The potential of smart speakers like Amazon Echo, Google Home, or Apple HomePod as a platform for games had not yet been recognised in my eyes at that time. I recognised the major advantage that operation by voice command offers in that it has few barriers to entry and is easy to master for people of all ages. So, what if smart speakers were used for adventure games or interactive detective stories? And what would such voice-controlled games look like in concrete terms?

When I told Figge about my idea, he was immediately excited. We imagined the Voice Assistant as a fairy tale uncle and the Smart Speaker like a campfire around which the whole family or friends gather and listen to stories together. But just listening is not enough, it has to be interactive! Perhaps the best metaphor for our idea is more like a role-playing game, where Alexa takes on the role of the game master and describes the world to the players and lays out the possibilities.

These were the basic considerations and so our journey began. It was adventurous and took us to places we had not known before. There were a few obstacles we had to clear out of the way, but in the end – spoiler alert! - we arrived at our destination.

From developers to storytellers

To be able to immerse oneself in narratives and control the course of action, one needs not only a gripping combination of text, language and images, but also a detailed game structure. Figge suggested bringing Hauke Gerdes on board for this. Hauke is a seasoned professional when it comes to Pen&Paper. They design and moderate for Rocket Beans TV, amongst others, and write top-quality stories. He was immediately enthusiastic about our idea.

Together, we developed an interactive story in which the players are gradually introduced to the theme and are asked to choose between different options at certain key points. The principle is called "CYOA"- the acronym stands for Choose your own adventure and refers to a game principle that emerged at the end of the 1970s in the form of gamebooks.
These gamebooks are divided into numbered sections. At the end of each section, a list of options is presented that refer to other numbered sections. You choose an option and skip to the page where that particular storyline continues. In principle, you can imagine the plot as a branching tree structure.

We keep this core principle, except not on paper. Instead, you are presented with your decision-making options in an app and accompanying audio from a smart speaker.

Think Ink: Scripting with open source code

We decided to include not only multiple narrative paths but also different beginning and end points in our game structure to make the game experience more individual and dynamic. This requires a lot of writing and game design skills, but Hauke did a great job. On top of that, we have also introduced a temporal dimension. This means that you get a different outcome depending on whether you choose an option at an earlier or later point in time. The result is numerous additional combinations, which further increase the replay value.

In our search for a markup language with which we could properly script the structure, we came across Ink. Ink is an open-source language and was developed by the company Inkle specifically for such branching structures (and their own CYOA games like the award-winning 80 Days). It was the perfect match and we fell in love instantly! The script in Ink is, with some minor adjustments, the heart of our technology and can be used by Alexa Skills.

A server as a converter

Despite all of this, the technical limitation of the skills was problematic. For example, Amazon has limited the number of MP3s that can be played to five. This is potentially disastrous for games that are largely based on audio and where 500 MP3s can quickly accumulate for an episode! In response, we built a server that we affectionately call the Voice Machine. The Voice Machine processes requests from the Alexa skill. For example, if 15 MP3s are to be played, it converts them into 5 equal parts or combinations. The MP3s are then automatically converted to mono quality, which is a handy side effect as stereo sound wouldn't be possible with a single Smart Speaker anyway.

This conversion happens on the fly, which means that nothing is stored in the permanent data memory. Instead, a hash value is generated that uniquely encodes each combination and can be used as a cache key. This allows the Voice Machine to fall back on local versions in the cache in the case of known combinations and only has to process requests that are new. Since the hash value is calculated from the HTTP request, we have placed Amazon CloudFront as the content delivery network in front of it and included the query parameters in the cache policy. In the end, each combination only ends up once in the Voice Machine, which saves time and money.

We make the Netflix for voice gaming

When the script was written, we thought about how to make the audio games available to players. The question was pertinent as browsing with smart speakers doesn't work very well. If you want to select a specific episode via the web, the voice assistant rattles off the whole list, which can be time-consuming. This led us to design not only the engine but also the content discovery.

We developed our own app similar to the Netflix platform. Through it, episodes can be selected and either sent to the smart speaker or played on the smartphone itself. The app was ready within half a year. We use Unity as our development environment, for which Ink conveniently provided a plug-in. Like we said, a perfect match!
Finally, it was time for content publishing, i.e. writing stories, setting them to music and releasing them. We brought on board a whole host of talented voice actors for the dubbing, including voices like Ingo Meß, who is known from Game One and Rocket Beans TV. All this was new, exciting and incredibly enriching for us.

We’re ready when the market is

Campfire is an experiment that has brought us an amazing amount of positive feedback. We made it to number 20 on the iOS charts without running any paid ads. Even investors were interested in us. There were a number of technical challenges that we grew from. We not only studied Alexa's skills in detail but also our own and became more aware of our strengths. It was specifically because interactive audiobooks are not classic mobile games, that we had so much fun developing them. Our technology works and has already been proven. We see great potential for the future and are ready when the market is!