Anyone know how to create own SAPI TTS voices? Are there any docs / apis / algorithms? I have not the slightest idea where I could begin. I want to use the voice in my own software and also I want to be able to sell it so it can be used in other 3rd party apps.
You need to make a component that implement TTSEngine API, see the documentation for details:
TTS Engine Vendor Porting Guide
For a practical example of the engine implementation you can download sample engine as well as you can check Espeak SAPI implementation.