William Carlos Williams clip-based text2image generation using Vadim Epstein implementation and Christopher Lee speech synthesis with vocodes.