Google has announced that DeepMind can mimic the human voice. While this feat might not seem that impressive given the fact that AI-driven voice synthesis has been around for some time, Google’s engine can mimic any type of human voice without little to no interference or input from the user.
We can all relate to Windows’ XP TTS engine, where the user could instruct the system to read aloud chunks of texts either in a female or male voice. Google’s DeepMind engine takes things to a whole new level.
By analyzing over 16,000 audio sample from seconds, ranging from people reading aloud new articles to intricate musical compositions, Google was successful in creating a system that can perform all of these actions without any help from the user.
DeepMind can mimic the human voice in detail, right up to inflections and even regional accents. This marvel of deep neural AI is called WaveNet, and apart from being capable of reading any type of text, it can even compose its own music.
Thanks to this new feature, WaveNet can now reproduce texts in English and Mandarin. The TTS predecessors of WaveNet used methods such as parametric or concatenative TTS to copy bits of texts.
As a result, the system was able to reproduce chunks of texts but was unable to figure out what to say after that. So, Google’s DeepMind AI instead of perusing through the text to copy them word by word learns from them and then computes a possible response.
Although the system is still in its infancy, it shows a lot of promise. Presently, DeepMind can mimic human voice and can create its very own musical compositions. Furthermore, the AI can even create strings of human-like voice without any textual support.
According to the team’s statement, the advanced TTS engine can currently render basic human-like sounds, but most of it is gibberish. Still, with a little bit of training, the revolutionary AI can potentially outsmart even the savviest speaker.
DeepMind can Mimic the human voice and create its own musical pieces. Of course, there is much to do before the engine will be fully capable of functioning on its own, but the overall results are more than encouraging.
Using this amazing new technology, computer scientists will be able to create advanced AI that can potentially voice out their ideas and interact with us on the same level as our peers would.
Image Source: Pixabay