Whoever said there weren't any good ones?

ID:275515
 
Mar 23 2004, 10:17 am
This is excellent. Maybe in future versions of BYONd we could use this in some of the coding instead of recording and players to interact with.

http://www.research.att.com/projects/tts/demo.html

If you haven't already seen it, tell me what you think of it.
Mar 23 2004, 11:04 am
mAYBE Its goOD FOR a synTHETIC VOice, but ITS not GOOD.!.
Mar 23 2004, 11:06 am
In response to Jotdaniel
lol, that took me a second to figure out :-p
Mar 23 2004, 11:09 am
Tseng True Guerilla wrote:
This is excellent. Maybe in future versions of BYONd we could use this in some of the coding instead of recording and players to interact with.

http://www.research.att.com/projects/tts/demo.html

If you haven't already seen it, tell me what you think of it.

"Excellent" is supposed to mean surpassingly good. The above isn't even ordinarily good.

We went over all that in a thread in Q&A recently. The state of voice synthesis technology completely blows, and is likely to for another decade or three.

Lummox JR
Mar 23 2004, 11:09 am
In response to Shun Di
At least you understood it. I accidentally had caps lock on when I started that post and just decided to roll with it.
Mar 23 2004, 11:13 am
cool link! and even included the ability to download the sound files!

if the DM language included a software-controlled polyphonic synthesizer, I know of a few algorithms taken out of the old Commodore 64's SAM program (Software Automatic Mouth) that would work out nicely for simple Text-to-speech.
Mar 23 2004, 4:54 pm
In response to Lummox JR
I figure the primary problem with voice synthesis is that computers just can't mimic the various inflections and exceptional cases which you learn as you speak English over your lifetime. The fact that it generates waveforms from samples instead of from actual tone generation also severely hinders the overall structure of the sentence.
Mar 23 2004, 4:55 pm
In response to Spuzzum
See my post.
Mar 23 2004, 5:01 pm
In response to Jotdaniel
Yours didn't say anything that mine did. Yours was just a joke about how bad they are.
Mar 23 2004, 5:07 pm
In response to Spuzzum
That was another joke relating to the aforementioned joke. Sorry to trouble you.
Mar 23 2004, 5:31 pm
Whoever said these aren't great obviously hasn't done this yet.

(Click the hyperlink on the page, if you didn't know what I meant...)


~Kujila
Mar 23 2004, 6:37 pm
In response to Spuzzum
Spuzzum wrote:
I figure the primary problem with voice synthesis is that computers just can't mimic the various inflections and exceptional cases which you learn as you speak English over your lifetime. The fact that it generates waveforms from samples instead of from actual tone generation also severely hinders the overall structure of the sentence.

Many formats of voice synthesis actually do use tone generation, including I believe the Bell Labs system. The real hindrance is that the human mouth and tongue and lips and teeth and larynx and nasal passages all form a complex instrument that is not adequately modeled or even understood by anyone.

Tone-based synthesis understands that voices have a certain base pitch combined with "formants", sharp dips in the frequency spectrum that tend to identify the speaker. Most systems model speech using the three most dominant formants, but human voices tend to have about six or seven as I recall. (Formants are likely one way we recognize each other's voices.) But the pitch and harmonics can change quite a bit during ordinary speech, and I don't think existing algorithms really do enough to cover that. This is of course to say nothing of the fact that our speech is a mix between voiced sounds and non-voiced.

Lummox JR