[Synth Home]

Speech Synthesis

Macintalk Text Assist MBROLA SAPI
Supposedly developed to aid the handicapped, I just think it's fun to play with.

Apple MacinTalk(PlainTalk) (3/21/2005) '040/PPC Macs with OS7.1 or better.

Mac OS X PlainTalk (3/20/2005) Speech Sythesis is still part of the Mac OS. The new OS X version boasts 22 voices where the old version had 34, now that's progress. Since checking it out would mean shelling out a couple grand for a new computer I don't think I will cover it.

Creative TextAssist (6/19/2001) For older Sound Blasters on Windows 3.x/9x

TextAssist 4.0 (1/12/2002) Creative stopped packaging TA with 3.0. Newer versions are available for purchase as a stand alone product @ www.TextAssist.com

MBROLA (3/21/2005) Very good Multi Platform, Multi-Lingual, speech synthesis package

SAPI(Microsoft Speech SDK) (1/11/2002) MicroSoft's contribution to speech synthesis

Microsoft Speech Server (3/21/2005) Microsoft is continuing to develope their speech technology but it's not available for Windows 98 so I'm out for now.

SayIt (4/26/2005) Looking for that Speak&Spell sound? Here ya go!

WinSpeech (1/17/2002) Not very good speech synthesis for Windows 3.x and Windows 9x/NT, but it has some cool features

Others? Yes there are many other speach synthesis engines. I choose the ones here because they meet my strict criteria:

  1. They didn't cost me anything. They either came with something I purchaseed or were available to download for free. I'm not a cheapskate, I just have limited resources.
  2. They work on systems I have. I have cutom built K6-2 based PC with win98, a 040 and a 601 mac with OS 7.6.1. Many packages are only avaiable for obscure systems like XP, NT, or OS X. Also Linux, and Linux is free, but it also takes a week to download on Dial-up.
  3. No assembly required, for the most part. Many of the groups dableing in speech synthesys must not want to deal with the general public or even non progamming audio engineers cause many only have C++ sources available.
  4. I could figure out how to install, use and have fun with the ones listed.

Types of Speech Synthesis

Sample Based

Non Sample Based

Speech Synthesis vs. Text to Speech

What to do with Speech Synthesis:
Aid the handicapped

Speech synthesis has been used to read text for blind people and for people who can't speak.

Automated Operators

I'm sure most of us have delt with these annoying non-people. I've heard a few that use Speech Synthesis most are use recordings of whole phrases and words.

Artistic

While the goal for the above is realism the same may not be said for Artistic applications. If you want a realistic voice in your music have someone sing it. If you want something that sounds artificial or alien Speech Synthesis may be what you want.

What it Lacks

What to Download and install (Many Huge Downloads): Sometimes it's a little difficult to know what to download to get speech synthesis to work. You have to download and install many related files before it will work and many more to obtain full functionality. Often you need an update for your operating system before it will work.

User Friendly: MBROLA and Microsoft SAPI are not very user friendly because they are not really for users but programmes.They come with source code and C++ examples and applications that aren't quite finished.

Music Friendly: Not at all. You can make some of them sing but it's not easy. There are no sequencers that support synth voices. If you want to use a synth voice in your musical creation you will most likely end up recording it to an external recorder and then dumping it back into your audio sequencer.

Process time: This is one thing that might be keeping people from developing a synth speech sequencer: There is almost always a significant delay in processing text in to speech, more so with higher quality synthesizers

Quality?: None of the Speech Synthesis I've heard sounds entirely realistic but that's what makes it fun.

Speech Recognition

The evil twin of Speech Synthesis, it's hard to find one without the other. I will be waisting this space here to say I won't be covering Speech Recognition unless I find a use for it. I sleep better at night knowing that my computer can't here all the bad things I say about it.