24/01/2013 - CereProc's Wonder Emporium of Text-to-Speech - Don't Get Mad, Get Snippy
But some emotions don't make a lot of noise. It's hard to hear pride. Caring is real faint - like a heartbeat. And pure love why, some days it's so quiet, you don't even know it's there.
Erma Bombeck (American Humorist)
We are a sensitive bunch - us humans - and this means that we all feel emotions and all react to the emotions we pick up from the sounds we hear in others' voices. For many reasons we tend to suppress the strong emotions we feel during normal communication, however we manage to make our underlying emotions known in a more discreet manner through small changes to speech.
Using our patented emotional speech synthesis technology, our text-to-speech voices achieve the minor changes that convey underlying emotions and display a human-like character. We can alter the underlying emotion in a voice and use this to simulate different emotional states. Take the sentence "I hope you had a nice day".
Here are three renditions of the same sentence using different voice qualities.
Positive but neutral: <usel genre='neutral'>I hope you had a nice day</usel>
Positive and very enthusiastic: <usel genre='calm'>I hope you had a nice day</usel>
Meaning YOU may have had a nice day, but I didn't: <usel genre='stressed'>I hope you had a nice day</usel>
These mild changes in the emotion of the speech are further enhanced if they are complemented by other cues. For example, below is an example using the Haptek talking head:
Need more detail? To read some more technical information on how we create emotional voices, have a look at the research and development page.
What about more extreme emotion? Sometimes we need to be able to express emotions fully. Although we can't synthesise extreme emotions, we can merge pre-recorded prompts with previously generated artificial speech output. In this way it's possible to pre-record high energy emotion and then synthesise dynamic content with less emotional charge.
Below is an example of this. Can you tell which phrases are pre-recorded and which are synthesised?
With speech synthesis we can convey very much more than just the text that is being rendered. In our next exhibit we look at the extremes of this. Producing synthesis where you can't understand the content but can experience the sense of communication: Beastly Antics.
In case you missed them, here's the Wonder Emporium Part 1 and Part 2 and Part 3.