Sleepwalking is the king of modern open-source [vocal-synthesis]. He founded the company Dreamtonics which builds the [synthesizer-v] software. Earlier in his career, he built the moresampler used by the [utau] community; he's the guy when it comes to open source vocal synthesis. He has a variety of open source libraries on his Github for various vocal synthesis, DSP, and audio processing use cases.
A notable library that he has created is [libllsm2] which is seemingly used internally by [synthesizer-v].
He put out a TED talk thing ~5 years ago going over his initial journey learning vocal synthesis. He talked about his initial experimentation in this space and some of his early projects which culminated in [rocaloid]. A few years after that, he created the initial release of [synthesizer-v] and eventually founded Dreamtonics.
It seems that his most recent area of focus is machine learning, using neural networks for [vocal-synthesis]. This makes sense since one of the new features of [synthesizer-v] is hybrid [concatenative-synthesis] and neural network-powered synthesis.