Archived

This topic is now archived and is closed to further replies.

bladesnet

does anyone know...

Recommended Posts

bladesnet    122
how could i alter voice thru DX8 in vb or even in c\c++ [8D] or could anyone tell me where could i find a good program whit what i could make realistic voices, because all i could find is CoolSpeach( and some others...) that have only some stupid voices like mary in a hall; mary in space; mary in....(and i''m just sick of it) and it sucks anyway please help[]

Share this post


Link to post
Share on other sites
xoxos    132
are you looking for voice alteration software or voice generating software? imho the at&t and bell labs demos, which can only be used on the internet, are the most realistic

http://www.research.att.com/~mjm/cgi-bin/ttsdemo

http://www.bell-labs.com/project/tts/voices.html

i use flinger (mbrola comparable if not better) which is a bit crappier for tts, but has the ability to sing, and thus be programmed precisely for timing and pitch.

for voice alteration, what you most likely want is a formant shifter, which makes the voice come from a seemingly larger or smaller body, but doesn''t change the pitch, so no darth vader/chipmunk as with a pitch shifter.

since i can''t program c++, my onlt attempts to construct a formant shifter have been with modular softwares, which inhibits the cpu performance, and thus the fidelity (hint programmers..) so all i have is a 1/2 assed vsti plug-in that will pitch formant up, ie. good for making male voices sound like squat females talking through their nose over a cheesy cellphone. this effect can be achieved by splitting a signal into low-pass and high-pass, feed both into a ring modulator, and adjusting the cutoffs and resonance. it takes a bit of fiddling but does produce a distinct pitched formant character. at least good for voicing goblins or something.

if you want to program it in c++, a channel formant shifter consists of two comb filters.. one analyses a signal for the amplitude at each frequency band, the other filter is uniformly shifted and each corresponding filter band is amplified by the output of the level detecting filter..

i''ve tried using offset delay times to generate the comb filtering effects, but think much better results would be achieved using this method with discrete filters. please pass any findings along formant shifting is still a high priced commodity, and it doesn''t have to be!

Share this post


Link to post
Share on other sites
Dalik    122
After I read the 2nd post something just came to my head that I thought was kinda funny, but pointless to post, but here it is anyways.

Does anyone find that they will change there design or an object(npc, gun etc.) based on there output on there creations? For example say you make a voice with sound software and you have this idea that this person has a deep male voice, but you cant seem to get that voice right when your making the sounds, so you change your idea of the person from a powerful deep voice to a female character that is sick. lol ..

Share this post


Link to post
Share on other sites