Voice generating device with MCU/DSP

I have a task to create a device that to reproduce a voice. The quality of the voice may not be good. It rather should be bad (having in mind that the device must cost efficient). Since I have non experience doing such devices I refer to you for some advises. I was googling last few hours to find how is it done. I found that there should be a microphone, low power amp, Pulse-code Modulator<--SPI-->MCU/DSP <--PWM-->Speaker (eventually some flash to store the sampled data).

The voice may have the quality of a cheap Chinese toy (this is the idea). So I think 4-bit PCM would do the job (I don't know - you say...). The song that the device will play is about 10 seconds long (not much of a flash I believe).

I will write the software. But, what ICs to choose, How much Flash should I provide. How many samples per second (approximately)?

I know it depends on the quality and the length of the song. If I achieve this quality I would be very happy:
https://www.youtube.com/watch?v=ybmEWLBkcsc
Is it possible to achieve this by using a MCU with 8-bit ADC and 5-bit DAC peripherals

Any references to a similar projects and any ideas are very much welcome. Thank you in advance.

I think that you could consider using the ISD1810 together with a small MCU.

For voice it is adequate to have 4k to 8k samples per second. (I think I once heard 8k is a speed used for audio streaming over the internet.)

Hence if you wish to store 10 seconds of sound, you'll need 40k to 80k bytes.

256 possible data values is satisfactory. This is convenient to have one byte per sample.

5-bit DAC playback... There is a chance that is satisfactory, but why not choose 8-bits?

Thank you guys. Thanks to you I found what I've been searching for. You helped me a lot.