Monday, April 12, 2010

Smooth Talkers: voice engines and text-to-speech software

I think a great many of you will agree with me when I say that voices are very important; they can give impressions of intelligence where it is not due, and give impression of stupidity where it is undeserved. When applied to the field of assistive technology, the above statement is even more salient. A good voice engine can make a B grade software look like a state of the art marvel. Where as a bad voice engine (like Microsoft Sam) can make a state of the art software look like something out of 1978 Battlestar Gallectica returning to haunt us.

Here is a news flash: most of companies producing text-to-speech software do not develop their own voice engine, that is because most voice engines are developed along the line of Microsoft SAPI 4 and SAPI 5, and it does not take more than 10 companies producing voice engines to saturate the market.

Thanks to my little obsession about making my over glorified Pentium 4 into a obedient wife with sexy voice telling me how much she worships me, read my books to me before I retire to bed, and proofread my writings;I am proud to announce that I have invested time and real money to test just about every voice engine on the market so my readers can just take my words for it.

Ivona Voice Engine http://www.ivona.com/helpVoice.php
Clarity: 5/5
Punctuation 3/5
Pronounciation 4/5
Compatibility 4.5/5
Overall 4.5/5.


AT&T Voice Engines http://www.naturalreaders.com/sample.htm
Clarity: 3/5
Punctuation 3.5/5
Pronounciation 3/5
Compatibility 4.5/5
Overall 3/5.

Realspek Voice Engine from Nuance . http://www.nextup.com/
Clarity: 3/5
Punctuation 4/5
Pronounciation 35/5
Compatibility 4.5/5
Overall 3.5/5.


Neospeech Voice engine http://www.naturalreaders.com/sample.htm
Clarity: 3/5
Punctuation 4/5
Pronounciation 3.5/5
Compatibility 4.5/5
Overall 3.5/5.

Acapella Voice engine. http://www.naturalreaders.com/sample.htm
Clarity: 3/5
Punctuation 4/5
Pronounciation 4/5
Compatibility 2/5
Overall 3.5/5.

Cepstral Voice Engine. http://www.nextup.com/
Clarity: 2.5/5
Punctuation 4/5
Pronounciation 2.5/5
Compatibility 4/5
Overall 3/5.

Acapella has the best quality second only to Ivona Voice Engine, but they tend to be software specific and will not work cross platform the way Ivona, Neospeech, and AT&T would.

What are your experiences with voice engines? Any pet peeves? Favorites? Questions? If so, post your thoughts in the comments section below.

Written by Luke Hsieh, Assistive Technology Advocate, Community Access Center, Riverside

The above article represents the opinion of the author and does not necessarily represent the views of the AT Network or its staff, although we must admit that we do enjoy Luke's writing style.

1 comment: