I asked him how he did it on Youtube & he said: "Speech is synthesized from subphonemes." ... "At all, there are 23 256-byte subphoneme waveforms, along with other 34 256-byte waveforms for instruments, that are well-packed." www.youtube.com/watch?v=-OKO...
We Are Vocoders by Dmitry_Milk at DiHalt2026 summer
YouTube video by Dmitry L
youtube.com