A tutorial for building a real-time conversational AI avatar that runs on a single GPU using LTX-2. The workflow combines a local language model, video generation, and audio synthesis — no cloud services required — so you can customize the avatar’s face, voice, and personality while keeping all processing and data local.