Command-line Windows 7 (64-bit) robot voice
April 9, 2011 3:08 PM   Subscribe

Seeking a command-line way to do text-to-speech in Windows 7 (64-bit), with a robot-sounding voice (something like the speech synthesizer in Wargames). I am able to do it with ptts.exe from Jampal, but the only voice I can use is Microsoft Anna. She is nice, but I want a voice that sounds like a robot. I am not interested in doing this in Linux at all.

I've done much searching, and found that it is tricky to add new voices to the Microsoft Speech control thing, apparently using 32-bit voices is a tricky operation and I couldn't get it to work. But this is kind of moot as I haven't been able to find a SAPI5-compatible robot voice to download. Other human voices are not suitable.

I am looking for something circa 1985-level technology, here. I do not need anything with intelligent inflection or tone or anything sophisticated like that.

I have found some software that would do robot voices from a GUI, but I need something that could be run from the command line (my goal is to call it from a Perl script with backticks ``).

It would be usable for my purposes if it can read the text to be spoken from the command line, a file, or both. The amount of text to be read would be small at any one time, just a few sentences or so typically, maybe going up to 1 or 2k in extreme cases.

I'd like the output to go straight to my audio out, but if it has to be saved as a sound file, I figure I could get some sort of Perl script to play the file, maybe? I hope that wouldn't be too hard but I haven't searched for modules that can do that yet.

Any help would be gratefully appreciated.
posted by marble to Computers & Internet (3 answers total) 5 users marked this as a favorite
 
Best answer: FreeTTS should work on Windows, and you could probably make Festival work if you install Cygwin. eSpeak has Windows binaries available. If you use the diphone-based voices, you'll get something that sounds at least moderately old-school (though perhaps not quite as old-school as you're looking for. eSpeak's formant synthesis sounds less natural, but I'm not sure if it's the sort of unnatural you're looking for.

If you're planning to drive this from Perl, you might want to look at the PerlSpeak module. Not sure how well that'll work on Windows, though.
posted by multics at 3:25 PM on April 9, 2011


You can also download DOSBox and look for abandonware versions of Dr. Sbaitso. IIRC he would say whatever you typed after "SAY". You can pipe DOSBox audio output to a WAV file.
posted by RobotVoodooPower at 7:07 PM on April 9, 2011


Response by poster: After poking around with multics's suggestions, I settled on eSpeak. It works great, sounds pretty darn robotic (definitely good enough for my purposes), and I tested calling it from a perl script and that went perfectly!

I have been poking around a bit tonight with the different variations of voices (it comes with plenty), and it looks like I could even set up my own quirky voice pretty easily (by changing some phonemes and other characteristics easily in a text file, and the procedure is well-documented).

Thanks a bunch, this is exactly what I needed!
posted by marble at 6:50 PM on April 10, 2011


« Older Great Dining for Less in RI?   |   How can something so cute make me this sad... Newer »
This thread is closed to new comments.