I'm looking at reasonably sampled music, say 11 or 22 kHz mono. With the current low prices of memory, you can store about 1.5 minutes of uncompressed waveform into a 8 Mb eeprom device. Do you think this will give better quality than audio stored into voice chips like ISD? Another possibility is to use MP3 chipsets - don't know about the price or quality impact of this choice.
From your site
http://www.romanblack.com/picsound.htm - you could get 33 secs of sound from a 512 kb eeprom. Than's not too bad. I have to check the quality of the sound, though.
Thanks for the ideas.