What an Indie Developer Learned from Publishing Google Assistant Apps – Marcellus Pelcher
Voicebot sat down with Marcellus Pelcher to learn more about his journey as a voice app developer on Google Assistant. His interest in voice and talking computers goes way back so the introduction of smart speakers was a lot like his childhood dreams becoming a reality. By chance, he found himself at Google when the latest era of the voice revolution took hold. However, at the time he was working in another Google division and that led him to launch several Google Assistant apps, many of them games, as an independent developer. Marcellus has learned a lot along the way and we asked him to share some of insights with Voicebot readers.
1) How did you get involved in voice app development?
Marcellus Pelcher: After creating five voice applications on my own, I wanted to work on it full time, so I’m now focused on voice application development within Google.
2) Why voice? Why now?
Pelcher: My family’s first computer was purchased when I was about four years old. I remember waking up to a machine on a desk and asking my mom, “Is it a TV?” She said, “No, it’s a computer.” I then asked if it could talk. She told me my brother and dad made it talk the night before. I found out that there was a program that you could load that said: “Help, I am stuck in a computer” in a very low bit rate voice. Also, my brother had Madden and Joe Montana games that talked. I loved hearing the computer talk!
Watching Star Trek, I wanted to make the science fiction I saw around voice a reality. One of the characters, Geordi, asked the computer to play music, to play a game, and to give him information about his trip. I specifically wanted to make that use case reality! Let’s fast forward to when the Amazon Echo came out in 2014. When I saw the device, I immediately bought it while it was in beta. I clearly saw the potential of the technology. I wanted to create apps for the Amazon Echo but was busy at the time since I just started working at Google. When the Google Home with the SDK released a few months later, I immediately took the opportunity to develop Assistant apps. I wanted to be part of the first generation of developers for smart speakers.
3) What was your first voice project and what did you learn from it?
Since I used to play chess a lot when I was younger, I wanted to put Voice Chess on the Assistant. I thought I would build a simpler app first so I can learn the platform and how the user and the voice bot would take turns during a game. So I started with Voice Tic Tac Toe. I was able to get Tic Tac Toe to be usable and then I moved on to Chess. But I had massive problems with voice recognition. Some I was able to overcome, but other proved too difficult with my time constraints. Easy problems like ‘Knight b4’ being mistaken as ‘Night before’ was solvable. Hard problems like the recognizer recognizing multiple squares as the same square, was difficult to overcome. For example, it would be something like both ‘a6’ and ‘h6’ mapping to ‘ASIC.’ I was trying to hunt down all of those issues, but ran out of time during my winter break.
4) What are some things you have learned from your other voice app projects that stand out?
Launch and then iterate! With graphical user interfaces, there are visual cues to what the user is allowed to do. They are constrained to the menus and buttons that you provide. With voice, they can say just about anything. Most users don’t read or listen to your instructions, so you have to see what most people do and provide a paved experience for what people say naturally. For the first few days after a voice app launching, I am glued to Dashbot transcripts.
5) As a voice app developer, how are you thinking about multi-modal development after starting with voice-only apps?
The way I am approaching this is to be voice-first and then build user interface around that. I think there is a trap to building a GUI and then building a voice interface on top of the GUI. If you build the GUI first, you may find yourself in situations where a certain flow does not lend itself to voice. But I believe going from voice and then adding a GUI you tend not to fall into such traps.
6) What is the biggest challenge that voice app developers face today?
Discovery and stable monetization. Developers need to know the inner working of how payouts are calculated.
7) What do you most hope to see from the voice platforms in 2018?
Some form of clear and stable way of making income.
8 ) What is your favorite voice app or the one you use most frequently today?
It is more a feature. I like to have the Assistant to ring my phone so that I can find it. As for voice apps, I love Akinator. It is the best voice app I have seen.
Check out some of Marcellus Pelcher’s Google Assistant apps such as Voice Metronome, Number Recall, Voice Tic Tac Toe, and State Capital Quiz. Also, Mr. Pelcher provides the code from many of his voice games for use by other developers on GitHub in the repo SysCoder.