Original text removed. Firstly let us sort out what people mean by quality. Sounds better verses sounds nicer? Nicer by whose definition? I personally find the festival voices bloody awful and the internation unpleasant. At high speeds they become unintelligible and they take a lot of processer time to synthesize speech. This is only my personal preference however. I don't want my computer to sound like a person; after all it is my computer synthesizing speech, not my wife reading to me ;-) I don't want my computer sounding like the star trek computer; as the star trek computer takes all day to say what it means. Human speech is hard to understand at speeds >400 words per minute; synthesized speech such as that found in espeak seems to work far better at these speeds. I have the same complaint regarding apple's voices; they sound natural but are barely understandable at high speeds and perform sluggishly. I am a long time user of hardware speech; accent, artic transport, doubletalk etc. I find software speech performs sluggishly in comparison especially a system like festival that seems loaded down with so much extra functionality. Certainly, festival is a flexible and configurable system; but I have no desire to learn scheme to read my mail, and the disk space footprint for festival is quite large. The higher quality the voices; the more disk space used and the more data needs to go to the soundcard. Just because I have a 2ghz processer does not mean I want to use a lion's share of it to synthesize speech. I tend to find the lag time between an application sending speech to the synthesizer setup and the speech beeing synthesized annoying on most systems, Jaws, Windoweyes and hardware speech responding as fast as i'd like. I find espeak responds quickly, and the speech is tolerable to listen to once you get used to it. Initially the default british english has far too much top end for my ears to handle, and it is so very loud. As someone who has used the echo gp and old school speech; I am perhaps more tollerant than most regarding quality. I'd much rather use espeak on the mac rather than the slow built-in voices such as fred and alex. Cepstral sounds nice; but still takes an inordinate amount of time to synthesize. we're working at getting hardware speech on the mac, and think that it will greatly increase the responsiveness of voiceover. Our doubletalk box will have open specs and will work on USB and serial, making it an alternative for modern hardware speech. Regards, Kerry.
