Visatronic: A Multimodal Decoder-Only Model for Speech Synthesis - View it on GitHub