OuteTTS, a novel TTS model, uses pure language modeling on LLaMa architecture (Oute3-350M-DEV base). It shows quality speech synthesis via crafted prompts & audio tokens, without external adapters or complex setups.