Voice Software: Selecting a Text to Speech Voice
Text to speech software reads text on a computer aloud, turning words into sound. By selecting a good text to speech voice (also known as "voice fonts" or "voice software"), you can improve the listening experience.
Many words can be said in several different ways. When deciding how to pronounce a word, Text Speaker uses a Voice. Voices describe how to say words, how to vocalize sounds, and how to guess the pronunciation of unknown words from the spelling. There are different types of voices, and voices designed for particular purposes. This article will help you select the right voices for your own use.
High-quality voices are made by recording "samples" of an actual human voice saying every sound that is part of the language. These samples are combined into a single file, along with instructions on pronouncing the letters and punctuation marks. When Text Speaker reads a word, the word is broken down into what are called "phonemes", individual sounds. For instance, DeskShare consists of the phonemes "d eh s k sh ae r". The way each sound is pronounced depends on the voice selected.
Text Speaker can also modify the way a voice speaks, by speeding up or slowing down the speech, changing the pitch, and changing the volume.
Voices can be divided into several groups:
Most text to speech software uses Microsoft's Speech Application Programming Interface (SAPI). Version 5 of SAPI includes several improvements that make SAPI 5 voices sound much better than SAPI 4 or older voices. Because of this, we recommend that you use SAPI 5 voices if they are available.
Words are pronounced differently depending on which language is in use. Also, some letters and letter combinations are pronounced differently. You should always use a voice designed for the language that is being read. For instance, we recommend using Alberto or Rosa to speak Spanish, and Alain or Juliette to speak French.
Even when people speak the same language, people from different areas pronounce words differently. For instance, a person from Boston might say the word "cart" the same way as "caught". A person from Arizona would pronounce the "r". Text to speech voices are designed to speak words with a particular accent, and you can select whatever accent is right for your use. AT&T's Mike speaks with a midwestern American accent, for example. Charles speaks with an upper-class English accent.
Some voices simply sound better than others. You can listen to the samples of High Quality Voices and judge this for yourself.
Make a Choice
When you are choosing voices for your own use, first decide whether to use a male or female voice. Then, consider all the factors mentioned above. If your listeners are French Canadian, Arnaud is the obvious choice for a voice. If you expect mostly English-speaking American listeners, you have a broad choice of IVONA voices. IVONA is a multi-lingual speech synthesis system that offers high-quality human sounding voices. Listen to the samples and decide based on voice quality. We recommend Eric and Jennifer. All of these voices are very understandable and natural-sounding.
Picking the correct voice makes listening to Text Speaker more pleasant and easier to understand. Listen to the samples today!