Sometimes in microcontroller projects we would like to have sound output. This is handy when you can’t look at screen, but still need to get status messages or other info. For such tasks you can use voice synthesizer chips that are able to read text with one of robotic voices. But if your needs aren’t that high, you can go with simple wave player. Skinny Satan have shared his experiments with speech synthesis using Atmega128 microcontroller. He doesn’t do any fancy algorithms here, but rather plays low quality wave chunks to make a message.
He converts sound files in to 8-bit and 8kHz data blocks that can be stored inside MCU. Then he uses PWM output along with RC low pass filter to get voice on speaker. There can be several words stored in array where algorithm can select which ones to combine to get a sentence. You are only limited by MCU flash memory which in fact is pretty OK for simple applications.