Text-to-Speech from Google Cloud gets new languages and voices

The Text-to-Speech service on Google Cloud will be expanded with a number of new languages and new voice sounds. The expansion should make the service attractive to more customers, from more parts of the world.

The app simply works by entering written text, which the service then converts into speech. The result is an artificial voice that pronounces the text. Text-to-speech services can be used for automatic answering machines in call centres. Systems in consumer electronics or cars can also make use of artificial speech. Even podcasts or audio books could eventually roll out of service with realistic voices. The more advanced the text-to-speech becomes, the more use cases appear.

According to SiliconAngle, twelve new languages or language variants will be added to the service. The additional languages are Chinese (Mandarin), English (India), Filipino, Finnish, Greek, Hindi, Hungarian, Indonesian, Modern Standard Arabic, Norwegian, Czech and Vietnamese. In addition, 76 new voice sounds have been added to the spectrum of the service. As a result, a total of 187 votes are now available in the service.

Realistic voice with WaveNet

38 of these voices were created by WaveNet, a Google neural-networks service that can generate realistic voice sounds. The total number of WaveNet votes is now 95. According to Google, these voices in particular can be useful when customers are in contact with call centres or voice control of devices.

Google’s product manager Dan Aharon states that the WaveNet voices are pretty close to real voices: “The practical effect is that for most listeners, a WaveNet voice makes the human-computer interaction a smooth and familiar experience. With these updates, Cloud Text-to-Speech developers can now reach millions of people in many countries with their applications – and with many more languages in the future.”