1. RE: TTS
- Posted by sephiroth _ <euman2376 at yahoo.com> Aug 03, 2001
- 436 views
gertie at ad-tek.net wrote: > Can you blend the sylables together during speech, or pre-blend into a > file, and then > play the file? right now i just pieced together a little program to test how i sound with the syllables i did record. it sounds bad :) i'll try to clean up a little, but i'm pretty sure you will be able to
2. RE: TTS
- Posted by sephiroth _ <euman2376 at yahoo.com> Aug 03, 2001
- 435 views
i'm not sure how anybody would feel about downloading a 4G file just for a simple TTS library, considering the largest eu file they ever had to deal with was exotica(or was it lemonheadz?) ;) mtsreborn_again at yahoo.com wrote: > PS. Don't work with syllables. Go the L&H RealSpeak > way and record full words, and later on merge them > together to form sentances. Otherwise your TTS won't > ever sound "human" nor fluent and casual. Just a tip.
3. RE: TTS
- Posted by gertie at ad-tek.net Aug 03, 2001
- 444 views
On 3 Aug 2001, at 16:25, sephiroth _ wrote: > > i'm not sure how anybody would feel about downloading a 4G file just for > a simple TTS library, considering the largest eu file they ever had to > deal with was exotica(or was it lemonheadz?) ;) Prolly larger, since you'd haveto record all manner of pronunciations, inflections, and voices. Plus the data indexing the 100 or so recordings of each word in every human language. I'd record data on the freq spectrum of phenomes (relative dimensions of each freq, etc), for phemomes. This way, they can be tuned for playback by pitch, harmonics, echo, and blending for playback as parts of words, no matter what the spoken language. Kat, thinking back to the discrete processing board she came close to making, using a hundred LM566 osc chips and even more tuneable-tracking phaselocked bootstrapped twin-t notch filters. > mtsreborn_again at yahoo.com wrote: > > PS. Don't work with syllables. Go the L&H RealSpeak > > way and record full words, and later on merge them > > together to form sentances. Otherwise your TTS won't > > ever sound "human" nor fluent and casual. Just a tip.