<...> but codec IC may hard to buy here (in Thailand :) ) <...>
Such difficulty looks strange, but if it's true I have couple of fun thoughts. You can try
1) codec chip from old PC motherboad from trash,
2) small cheap USB Audio Adapter, that should be in your local shops.
Maybe you can buy suitable STM32 kit with audio interface like this STM32F401C-DISCO
http://www.st.com/web/catalog/tools/FM116/SC959/SS1532/LN1848/PF259098.
For DAC, i still thinking to use PWM techniuqe with lowpass filter to regenerted sond, will try this weekend and update again.
Sound quality that is enough for "Uaaa-Uaaa" inside plastic doll will disappoint you.
My advice - use the way CocaCola suggested. At least for the beginning.
But if you have strong attraction to digital approach and want to make your own effects then I'd better start with PC application that captures sound from sound card, processes it and playbacks through the same sound card.
I am very interested in your report about reached quality by PWM+LPF solution.
Here is one demonstration "Arduino playing Audio with PWM DAC"
http://www.youtube.com/watch?v=XLm-QY0Bmno