“ElevenLabs' co-founder, Piotr, developed a new approach for creating voice models by applying ideas from transformer and diffusion models to the speech domain.”