 |
 |
 |
 |
 |
 |
 |
Forums - Music techology
Subject: Vocaloid
Message 11/11 20-Feb-04 @ 03:32 PM - RE: Vocaloid
we gotta sell a lotta copies to make this thing pay :p
that first 'check it' demo.. the last 'leon and lola' might have been purposefully articulated to sound like an old tts.. if not, then it doesn't sound like much more than a higher sampling rate version of flinger (if you haven't heard flinger, check my panicnow page for 'virtual paradise' or 'dawn.') which, btw, is freeware..
mbrola is also freeware.. dunno what sample rates it supports (flinger is only ~16kHz but.. works) but importing your own voice is possible on both platforms..
jic anyone's reading this that hasn't followed my 'saga of amateur freewareist exploits into speech synthesis,' here are a few leads to getting.. equally interesting vocal processes to the vocaloid..
mda talkbox.. there is no description, but i reckon it's a vocoder that uses fft instead of a number of bands.. feed it with a bright sawtooth and the output is almost like the original only with a bit of the characteristic odd harmonics prevalent, like the original only a bit flat/digitalised, prolly about the same quality as the cheaper autotuners..
haven't tried it yet, but add some high passed white noise (@~8k?) might save some of the 'sparkle' of the unprocessed vocal..
..i built an engine for my q-series algorithmic synth generators (yeah, me me me.. yawn..) for randomised vocoding, ie. the plugin changed pitch and stuff to the tempo.. i included a few modulation features, ie. a vibrato lfo that would fade in and out, a glide function that would modulate.. all in all i think only 3 different modulations to this signal, and the results were pretty cool for organic emulation.
in other words, an engine with perhaps 5 or 6 modulators (ie. 3 lfos, varying glide et c.) should produce a human, rawish sounding sequence.. so if one were to build a special synth in synthedit with these modulations configured to your control preferences..
..if you're familiar with the random vocals for my algorithmic bits, the major fault of that design is that there is no amplitude modulaiton of the control signal.. so you get vibrato, but no tremolo, or varying levels, which is very unnatural for voice.. consider that :p
..and if you're just looknig for a goofy sequenceable synth vocalist, there is a free called synger on my homepage.
options to the new dance craze.
that first 'check it' demo.. the last 'leon and lola' might have been purposefully articulated to sound like an old tts.. if not, then it doesn't sound like much more than a higher sampling rate version of flinger (if you haven't heard flinger, check my panicnow page for 'virtual paradise' or 'dawn.') which, btw, is freeware..
mbrola is also freeware.. dunno what sample rates it supports (flinger is only ~16kHz but.. works) but importing your own voice is possible on both platforms..
jic anyone's reading this that hasn't followed my 'saga of amateur freewareist exploits into speech synthesis,' here are a few leads to getting.. equally interesting vocal processes to the vocaloid..
mda talkbox.. there is no description, but i reckon it's a vocoder that uses fft instead of a number of bands.. feed it with a bright sawtooth and the output is almost like the original only with a bit of the characteristic odd harmonics prevalent, like the original only a bit flat/digitalised, prolly about the same quality as the cheaper autotuners..
haven't tried it yet, but add some high passed white noise (@~8k?) might save some of the 'sparkle' of the unprocessed vocal..
..i built an engine for my q-series algorithmic synth generators (yeah, me me me.. yawn..) for randomised vocoding, ie. the plugin changed pitch and stuff to the tempo.. i included a few modulation features, ie. a vibrato lfo that would fade in and out, a glide function that would modulate.. all in all i think only 3 different modulations to this signal, and the results were pretty cool for organic emulation.
in other words, an engine with perhaps 5 or 6 modulators (ie. 3 lfos, varying glide et c.) should produce a human, rawish sounding sequence.. so if one were to build a special synth in synthedit with these modulations configured to your control preferences..
..if you're familiar with the random vocals for my algorithmic bits, the major fault of that design is that there is no amplitude modulaiton of the control signal.. so you get vibrato, but no tremolo, or varying levels, which is very unnatural for voice.. consider that :p
..and if you're just looknig for a goofy sequenceable synth vocalist, there is a free called synger on my homepage.
options to the new dance craze.
Pages: 1 2
There are 11 total messages for this topic
Reply to Thread
You need to register/login to use the forum.
Click here to Signup or Login !
[you'll be brought right back to this point after signing up]
Back to Forum
