Category: Geeks r Us
Could you share with us what you think of these voices? What kind of memory do they use? Do they work well with screen readers?
I like how Chrystel sounds on the demo, but I cannot speed her up via the demo. I'm curious how she does at a faster rate.
Thanks
They are quite ram intensive and at least 1gb is recommended. They still are quite passable at faster speeds.
Cheers
agreed with sjtailor here, ram intensive, but workth the ram.
What alternatives do you suggest? Something with a small footprint?
I personally like the NeoSpeech voices or many of the ScanSoft voices, which can be bought from Nextup.com and possibly other places as well. Only problem is that to buy from Nexup you have to be a registered user of their Text Aloud program or the voices won't install. The only exceptions are the aforementioned NeoSpeech voices. The NeoSpeech voices sound, to me at least, more natural than the so-called Natural Voices. This is quite the irony since the NeoSpeech voices are totally synthetic whereas the Natural Voices are based on actual recorded human speech. The problem you run into with that is that when the voice comes across a word that's not in its dictionary it really mangles it. And you'd be surprised how many common words that can include.
try akapella voices too from nextup, they are pretty good also.
Here's what I use. For things I want to read, like the few times when I've had to use speech for textbooks and what not, I slow down the ATNT voices and they work fine. I don't think they sound all that weel when sped up, so I use eloquince for that. Neospeech Kate sounds alright to a point when sped up, but begins to be distorted when sped up too much.
If you slow Neospeech down to a very slow rate of speed, they sound very drunk. If you find yourself with way too much time on your hands, try typing in a string of profanity, and you'll think you've run into a street person on something.
Lou
How about the
Cepstral voices? I think I spelled that correctly. Are they a small footprint, use little ram?
They sound like they... well... have had things done to them. For example, Amy is blowing away in the wind, Laurence is mentally impaired, and david's trying to sound like a robot. Also, neospeech is based on recordings of human speech. The engines just use different algorithms for concatination of the phones and diphones in the databases. Neospeech's is less intensive, and doesn't take too much space, while at and t uses a more ram-intensive algorithm. I recommend them for a natural voice that is ok in responsibility, but use sapi4 whenever possible, not sapi5. By the way, I have a copy of Mike I can give you. If anyone wants, please pm me. I don't know whether or not it'll fit on my dropbox, but... we'll see if someone wants it, won't we?
Isn't that illegal?
Ok, then I'll throw out this CD. I didn't install it, and I can't get the actual disc to someone. I was just suggesting if someone wanted the files I could send them over the internet. Better not break any laws.
The reason, by the way, that I didn't install it is that I use the real speak direct voices for JAWS. That, and the supposedly illegal eloquence on NVDA when I goout.
Oh, and I used to have a copy on my old computer, that's how I know how well it works.
Ok, I have found someone to give the CD to, offer closed.