Here are a couple of ongoing projects in the Vocal Synth community. If you're interested in working on any of these projects please contact us.
A formant based speech synthesizer based on eSpeak, which was started by Jonathan Duddington. It was later forked into eSpeak NG, which supports many more languages.
Reference implementation of the MBROLA (Multi-Band Resynthesis OverLap Add) algorithm for speech (and singing) waveform generation. Although it is diphone-based, the quality of MBROLA's synthesis is considered to be higher than that of most diphone synthesisers released in the 90s.
The Festival Speech Synthesis system is the classic free software one that runs on GNU/Linux. For long time it was the best option for building synthetic voices (including singing ones).
The WORLD vocoder is a high-quality speech analysis/synthesis system, that is often used in singing synthesizers including UTAU and NNSVS. It is used by CeVIO (a part of analysis) and OpenUTAU (as a resampler backend).
ENUNU and NNSVS are two alternative backends for UTAU. While UTAU uses concatenative voicebanks be default, thanks to the plugin ENUNU, it can be also used with the Neural network-based singing voice synthesis library NNSVS.