Alexa, sing Happy Birthday

About a month ago, I learned about this easter egg on the Echo. Ask Alexa to sing Happy Birthday and she’ll sing. Of course, it’s not really Alexa — it’s the voice actress whose voice they synthesized for the TTS voice singing a recording of Happy Birthday.

However, this singing of Happy Birthday is a great preview of things to come. It’s probably how the TTS voices are going to sound in the next two years.

Voice synthesis is one of those areas where there could be a Turing-like test. Can it fool us into believing it’s an actual person speaking?

A few years ago, we had a big internal debate about which TTS voice to use on the Ubi— whether to go with a paid voice from Ivona (owned by Amazon) or to use the free (and nasally) Android voice. In the end, we paid to use the voice because it was our “screen” to the user and was much smoother sounding.

One of the ways that we can start to get a more personal interaction with voice devices is for them to respond first by mirroring our own speech — either with the same inflections or cadences. For example, if someone speaks quickly to a device, we can detect word rate and then change the STT speed. The next step would be to change the inflections and eventually add some type of emotional flare.

It’s likely that we’ll get awesomely realistic TTS voices very soon. The next step will then for them to pass this new Turing test.