And, by a lot, I mean, holy cow, A LOT. We undertook this journey to build an app because the research that we did to find a solution came up dry. We looked for something that would allow David to continue to use his own voice. And our search didn’t lead us to find something that had the features we felt were needed at the level of quality necessary. Now, there may very well be something out there that does.
There are a lot of great apps out there, and the internet is vast. But, when you’re in a race against time, you do your best to find what you can as quickly as you can.You have to make some decisions quickly. David has spent his entire career saying, “let’s not let perfect be the enemy of good.” This was his a way of encouraging clients to make a decision to move in a direction. And, that’s what we did here. We moved.
Pick a direction and go
The direction we picked was to make something to take maximal advantage of the capabilities of ElevenLabs voices and voice clone technology. This was something that we didn’t find in our rapid initial search. The MVP proved that a simple, elegant interface could render quality voice output with a simple text message quickly and easily on device.
With that proof of concept in hand, literally, the vision was clear. Make it scalable, secure and robust. And, make it have a little bit of personality. And with characters like these, I think the personality comes through.
Since our initial search, we’ve come across a few other solutions that do connect with ElevenLabs, but they lack some of what we’re now trying to do.
But, it has to be a little bit serious, right?
Yes, it does have to be more than a little bit serious. Having never built anything that was scalable, secure and robust before, we did spend quite a bit of time working through those details. Thank goodness for Google (and YouTube, and Microsoft’s Copilot, and…), but mostly thank goodness for Google which is where the back-end of the app is built.
The entire infrastructure is built in the Google cloud. Google handles user management, storage, networking, scaling and all the bells, whistles and occasional goose honks necessary to turn text into speech with the click of a button. Along the way, we bolted lots of different building blocks onto the fuselage, but the plane flies pretty well, actually. We continue to modify it. When we introduced Merlin, we even added a whole new set of avionics to the cockpit.
What about those avionics?
Yeah, what about Merlin? Well, Merlin takes advantage of existing OpenAI advancements in artificial intelligence and some carefully crafted prompt engineering to assemble suggestions for text blocks that can mitigate the need for typing on your part. With a few button clicks and a little bit of direction, Merlin waves his wand and generates three options for you to use or edit (on the iPad), or regenerate with different options. David was so excited about it, he was doing the happy dance for days. By the way, nobody ever needs to see the happy dance.
Merlin does not yet have a brain, though. He can’t remember what he had for breakfast this morning let alone what you asked him for a moment ago. We’re working on that. When Merlin gets a brain, he should create much more personalized text suggestions that will be increasingly relevant to you. We think that he will reflect your vocabulary, your verbal style, and your context. This is going to take some time though. So, be patient.
Where are we headed from here?
Right now, Talk to Me, Goose! is out in the world, and we’re gathering feedback on how people are using it, a little bit. Given that, we continue to make some tweaks to make it more resilient. Most of those have been behind the scenes. We improved the user experience. And, we cleaned up some of the things that we did in a hurry up front. We also fixed a number of things that we learned along the way could be done more efficiently or effectively.
We really hope for three things:
- To get this in the hands of people who really need it
- To get the costs to serve it down as much as possible
- To reduce the costs for people who need to rely on it – to make it free, if possible.
Simply put, we want to continue to make it better. To that end, David is meeting with anyone who wants to talk seriously about ways to do that. If you’re that person, reach out on the Contact form and get in touch.
Until then…happy flying…go check it out, and give it a test flight.