There are two ways of giving voice to your microcontroller design: one is to use speech synthesizer and another play recorded sounds. Synthesized voice is far from natural and so may seem to cold in many applications. It is algorithm based method and so doesn’t require lots of storage memory to get long speeches. Recorded voice is still better choice when you need natural sound. Problem is that you need to record first in order to play back. Boris Landoni designed great Arduino voice shield which uses ISD1790PY chip.
Chip is quite robust and actually can be used as standalone solution or interfaced to microcontroller using SPI interface. Chip can store up to 120s of recorded sound. It already has a microphone where you can record messages directly in to chip. Chip also have built in audio amplifier so only speaker is needed to be connected using audio jack. There is also a special software where messages can be uploaded from computer and organized. With this Arduino voice shield you can build lots of fun things starting with various home automation solutions, security systems or simply data output using voice messages.