Welcome to part two of the Why Choose LumenVox series here at LumenVox.com. We will discuss why you should pick LumenVox versus one of our competitors to be your speech recognition provider. You'll remember in part one we covered some of the distinct differences in the LumenVox philosophy, how we want to be your partner, not your competitor, and how that differs from some of the other speech vendors out there.
In this video, I am going to tell you how this philosophy translates into our actual business model, how we sell licenses, and also some of the technology we give you and what you get when you pay for licenses.
To get started, the basic bottom line thing is cost. It's a big differentiator because our competitors are about two or three times the price of LumenVox in most situations. So that means they better be giving a whole heck of a lot more if they're going to charge you two or three times more, and we just don't think that's the case.
We talked about in part one how accuracy is pretty similar across the board, so if you're getting roughly the same accuracy, why would you pay two or three times the price? The fact is they're giving less, because we give you a lot more that we will talk about in just a moment in regards to the technology, some extra features we include, and we charge less for it. Therefore, when you pick our competitors you're paying more for less product — that's a bad business decision.
LumenVox licensing is simple and straightforward. We have two license types and each license is going to give you a single interaction with the Speech Engine. So if you have one person talking to the Speech Engine, you use one license. If you have five people talking to it, you need five licenses, and if one person hangs up that one license goes back to the pool, pretty straightforward.
We have two tiers. One is what we call our lite license and it limits the number of words that can be active at a time. This means it's fine if you have a small application. It doesn't need a whole bunch of commands at any given prompt and it could be five hundred words at one prompt and a different five hundred words at the next.
The full license is unrestricted up to the point of what's technically feasible and that changes a bit. The number of words that are possible with a full license depends on your computer hardware. Right now, it's in the tens of thousands of words that can be recognized with a full license. We would eventually like to have it up in the hundreds of thousands of words for you.
The license pricing is pretty straightforward. The full license cost has a retail price of $595 and the lite license costs a retail price of $245. These are both retail prices, if you're buying as a reseller, buying in bulk, you can talk to us about volume pricing.
If you want, there is an optional licensing fee you can pay per year after the first year that gives you access to upgrades. We put up new updates every year with better accuracy, new features, better performance, and it's about fifteen percent per year of the initial licensing fee. This is a good value considering we've been told that our competitors charge as much for their maintenance as we do for our initial port.
Therefore, buying one license from LumenVox is about the equivalent of paying one year of updates with the competitor and our updates are just fifteen percent per year. This means after the first year you're saving 85% every single year. What's that do for your total cost of ownership for the LumenVox speech solution versus some of our competitors?
Our licensing system isn't set in stone. Like everything at LumenVox, we're flexible about this. We're a small company and we like to think of our selves as agile, adaptable and flexible.
So if you need something different than our licensing system provides, we can help. Let's say you need pay per decode, you have some burst demands, some peak volume, then it goes down and you want to pay for every time you use it rather than just a flat fee.
Talk to us and we can work with that or a subscription model, maybe you want to pay every month or every quarter, let us know. Talk to business development and we can work something out for you. We will build a custom solution for you if that's what it takes — a business model that fits both of our needs and I'm confident we can do that for you.
In terms of technology, we do provide quite a bit more than our competitors as well. Included with the license cost all of our languages that we support. Therefore, for that single license price, you get every language. That means you get U.S. English, U.K. English, Australian English, Mexican Spanish, Colombian Spanish, and you also get Canadian French.
That's a good mixture of languages that cover all of America and other parts of the world for a single price. This means if you have an application that's successful in the U.K., you can bring it over to North America where you can turn on U.S. English, Mexican Spanish, and Canadian French. Now you've got, for the same licensing cost, you've just moved it to a different continent and supporting a much broader base of users.
Your applications can grow without your budgets needing to grow and that's a really key thing we think.
In fact we are all about helping you scale. Our entire architecture and infrastructure is designed from the ground up to help you change the size of your application. We use a client-server architecture that's sophisticated. We have a very thin client that talks to a process called a speech server and this speech server could exist on the same box as the speech client or it could exist on a different machine and they can talk over the network.
The speech client will traditionally reside on your application server or your voice platform, that kind of thing, and it sends audio and grammars to the speech server. The speech server does the heavy speech recognition and it's actually that speech server that takes up allot of CPU time and memory. Speech recognition is computationally expensive.
The cool thing is you don't necessarily want to bog down your voice platform with all your speech recognition requests so you move to a new box. Then your application is successful and it grows, if you need another speech recognition server you can just plug it right in or a dozen. The client is smart enough to know that I have list of speech servers to send speech requests to and it does some load balancing.
This way, no single server is overloaded or overburdened, which is really nice and also says if one of these guys goes down I can go ahead and not send them anything until it comes back up. You get redundancy, load balancing, scalability, all built right into the software.
If you wanted to, you could have yourself a New York office, a L.A. office, and they're going to talk to each other over the Internet and send requests. There's also an office in Tokyo and if its server failed it's automatically going to switch over to New York servers for a temporary time and all this kind of stuff. All that's included at the same price, complicated network infrastructure and distributed architecture is all there for you.
The final bit of technology we provide, which I think is the absolute coolest thing we have, is something called the Speech Tuner. It's an award-winning tool that our competition really doesn't have because I think in large part they want to sell you tuning as a professional service. They want to charge you after the fact, help you make your applications better.
We give you the tools to be successful so that you can get this kind of stuff done without paying an arm and a leg. What happens is that every time a caller calls into the system and says something and gets recognized, we save off what the caller said in a little file. With that, we save what the Engine recognized. Now you can use the Tuner to load up all these files, listen to them and hear what your callers are saying, and look and see what was right and what was wrong.
Once you start transcribing the audio, you get reports and you can see, "Ah, my average accuracy was this, my average confidence scores were these." You will get all sorts of neat reports back. The powerful thing comes once I have these reports and I've identified some problems, I can say, "This word here is constantly getting misrecognized. What can I do to fix that?" Maybe I need to alter my grammar, add in a different pronunciation or I need to change all my prompts and the grammars and add in new options. Maybe my volume is too low and I need to play with my Engine sensitivity settings so it will listen to this lower volume.
You can change all this kind of stuff inside the Tuner, and click a single button and run a new test. Now it's going to take this audio that we've already collected from our live production systems and it's going to run it through the Speech Engine again using the new grammars and the new settings. This process tells you now your accuracy is this, it was this, and here's the change and here's what is now right, now wrong and what the difference is. Therefore, you could test changes to a production system using real live data without altering that production system all from a development environment.
That's key because speech recognition is all about doing these kind of tuning fixes. These little tweaks to an application can get big performance increases. That's how Redmond and General Motors got that 42% bump in accuracy that we talked about last time. Because they did a lot of tuning, they worked with us, they used our Speech Tuner and we give this to you free of charge. You buy a license for the Engine, you get a Speech Tuner, and you're free to use it.
Okay that wraps up this section. In the next video, we're going to conclude our talks by looking at some customer testimonials and some of the other great resources that we make available for you for free because we want you to be successful. Thank you.
© 2016 LumenVox, LLC. All rights reserved.