Board Thread:Questions and Answers/@comment-19840316-20140512164337/@comment-5334617-20140519184801

Sounds are compression waves in the air. Compression waves are generated by speaker cones, moving forward (pushing the air out, or causing higher pressure) and backward (pulling the air in, causing lower pressure). If the speaker cone moves very fluidly, like a sine wave, the tone is very pure; if it just jerks back and forth between two positions, that's the sound of a classical electronic bleep; if it does more interesting things, it makes tones with different textures and characteristics.

In the built-in help manual for SmileBasic, chapter 44 Audio (Expert), there is a page on BGMPRG. There is a graph there, that shows 0 as a value representing the speaker cone in the middle of its range, the 'neutral' position, then moving forward (presumably) to a maximum, then backward to a minimum, then back to where it started (it always ends where it starts). The air pressure increases for 3 units of time, then decreases for 6 units of time, then increases for 3, per cycle of the sound. This example is only 24 hex digits long, and the system expects 64 or 128, so you can create more sophisticated shapes than the one shown.

You may be able to get some vowel sounds like 'ooo' and 'eee', since they're not too far from pure tones, and you can get 'shh' or something close to it... I suspect speech synthesis is a lot to ask, but I can't say for certain it's beyond possibility.