If you’ve read the links above, you know that here at My Voice we are not a manufacturer of high quality Text To Speech (TTS) voices. Rather, we provide features, UI and a great user experience around TTS voices provided by other parties.
So which are the best TTS engines with the best voices? Well, we’ll discuss that in another blog post shortly – but Google’s WaveNet voices feature way up near the top.
We want to know what YOU think about WaveNet, so please read below and head over to our polls on Facebook to offer your opinion – all polls are listed at the bottom of this post.
What is WaveNet
WaveNet is a collection of Text To Speech (TTS) voices available via Google’s Cloud platform. For the technically astute of our readers, you may be saying hold on one moment… Cloud?!
That’s right – WaveNet is not an on-device TTS engine, which is traditionally how devices synthesize text into speech. Instead, WaveNet does all synthesis in the cloud – so you might type something on your device, and an app would send that off to Google’s Cloud WaveNet service to be converted to speech, and this service would then send the completed speech back to your device. Pretty cool, right?
WaveNet uses the same technology used to produce speech for the Google Assistant, Google Search and Google Translate services – its arguably the most advanced TTS technology in the world currently, powered by machine learning and AI.
The speech that WaveNet produces is considered to sound more natural than other TTS engines can produce. Synthesized speech includes more human-like inflection and emphasis on syllables and words. That’s as deep as we’ll go in this post, but if you’re interested in reading more in depth about how WaveNet was created, check out this post by DeepMind here.
Sounds great! What’s the catch?
To put it bluntly, WaveNet voices are expensive.
As a developer, access to WaveNet synthesis is controlled by usage limits and cost per character limits. We get up to 1 million characters for free each month – a character refers to a single letter, space or symbol. After that, Google currently charges developers $16 per additional 1 million characters. So if you use 3 million characters in a month, you’re paying that twice.
Lets put that into perspective. My Voice currently helps users to speak nearly 1 million sentences each month – and a sentence could be anywhere from a couple to many thousand characters. We don’t track the number of characters spoken in our app, but you can imagine that it’s up in the tens of millions per month.
Sadly, that’s not a cost we can afford to pay – My Voice is run by a single developer as a passion project, and the income we make is absolutely minimal.
What options do we have?
There are a couple of ways we could integrate WaveNet voices into the My Voice app, but we need to be clear that the cost of this would have to be covered by users. We want to see what the appetite for this is, which is where you can help us out! Take a look at the options below, and leave a comment on this post if you have a preferred choice – or take part in our Facebook polls as and when they appear over the next few weeks!
Option 1: Subscription Model
Our first option is a monthly subscription model for access to WaveNet voices. This would be optional, and users would still be able to use the My Voice app for free with standard voices. We expect a subscription model would cost between $10-15 per month, and you would have unlimited usage of WaveNet voices.
Option 2: Single Purchase Model
Our second option would be a repeatable in-app purchase model – where users essentially purchase 1 million characters worth of WaveNet voice access. We would expect to provide this at cost value, so approximately $16 each time this is purchased. When you reach your limit, you could simply buy more characters, or continue to use the app with standard voices for free.
Option 3: Use your own API key
Our third option is one implemented by other apps such as Tasker. Its a more technical approach for users, and involves you signing up for your own cloud account and getting what’s known as an API key for WaveNet. You can imagine this as an access password. You would enter this into the My Voice app, and be able to use your own allowance (1 million per month free), after which you would be charged (by Google, not us) for any further usage.
This would enhance functionality for users, at no additional cost to us. However this option could likely confuse some users, as it’s not straightforward. We pride ourselves on the simplicity of the My Voice app, so obviously there are some concerns about this approach.
Have your say!
You can have your say on this discussion by leaving a comment at the bottom of this page, or answering our Facebook polls which are listed in order below. Thank you for your responses!