How are electronic musical greeting cards made?
These sound chips were and still are a creation of Seiko/Epson corporation or one of its subsidiaries. It was a spin off product from the extremely low power CMOS processes that they developed for the watch business. The design of the IC itself is actually a 4bit uProcessor with mask ROM for the different "tunes". They also provided more expensive OTP (EEPRom) based units for experimentation before changing the mask.
I'm sure that there are a number of companies that could make these and searches on "greeting card music chip" show lots of Shenzhen references.
They are purchased as KGD (Known Good Die) in waffle packs and then glued to the PWB (PCB), wired bonded to the pads and covered with epoxy. I'm sure now a days you could get them with bump bonds.
The volume of these cards was high enough that it probably justified a custom chip. The manufacturing process is called "chip on board", or often just COB. The bare silicon chip is mounted diretly to the circuit board, then a blob of epoxy put over it to protect it. In high volume this is cheaper than using a packaged chip.
If this was a custom chip, it probably contains a ROM and a little sequencer to read the sound samples, then probably a PWM generator to make class D audio open loop. Those are just guesses on my part, but would be the first approach I'd investigate if tasked to come up with something like this. Above all, it has to be cheap. Sound quality isn't much of a issue. The speaker has to be so small and cheap that it will be the limiting factor of quality. You want the sample rate just high enough to not add too much quantization noise. It only has to work for a few minutes, then it will get tossed anyway.
You will notice the card has four things.
1 small batteries
2 the black blob (the electronics)
3 a crude switch that slides between contacts
4 a cheap and nasty loudspeaker
The chip is switched on when power from the batteries is connected to it through the switch contacts.
The black blob is an epoxy resin covering a small integrated circuit chip connected to the PCB with wires. This is exactly the same type of thing you would normally expect in a 'normal' IC chip with legs. By just using the chip bonded to the board it is cheaper and smaller.
The chip doesn't require capacitors or xtals because it has an internal oscillator (just like many of its bigger brothers.)
It essentially divides this frequency and follows a simple program to output sound. By using PWM (pulse width modulation) it can create speech and other 'sounds'.