speech synthesis demo

Luciano Notarfrancesco luciano2 at mail.ru
Sat Jun 19 22:04:34 UTC 1999


> 
> Luciano -
> 
> You mentioned in the previous post about using Image Segments.
> Instead of disk, can Image segments easily be sent across the
> net from one image to another? Or is this the idea?
> How would it be done?
> 

Image segments can be sent across the net... but I am not using them in that way. Actually, it would be very slow doing that. The diphones (pairs of phonemes, sampled and LPC coded) are too big... the Kurt voice uses about 8.7 MBs, so instead of having them in memory all the time I store it in disk in image segment format, and every time a diphone is needed I load it.. and after a GC the diphone is lost and if you need it again it will be loaded again. This works because diphones never change, so I only store them once. Doing this across the net would be too slow, I beleave.

Regards,
Luciano.-





More information about the Squeak-dev mailing list