# Transformer Diffusion

Pixart-α is a Transformer-based text-to-image generation model capable of directly producing 1024-pixel images from text prompts, with significantly superior training efficiency compared to similar models.