Abstract: The generation of high - fidelity and complex text to image data has consistently posed a significant challenge. Although pre - trained autoregressive and diffusion models have achieved ...