Abstract—Text-to-Image (T2I) generation has advanced rapidly with the emergence of diffusion, transformer-based, and vision–language models. These approaches enable high-quality, semantically aligned image synthesis from textual prompts. In addition, zero-shot and few-shot adaptation strategies and parameter-efficient fine-tuning methods have improved the flexibility and scalability of T2I systems. However, these advancements have introduced significant computational and environmental challenges. In this paper, we present a structured review of T2I generation, focusing on the relationships between model architectures, adaptation strategies, computational efficiency and environmental sustainability. The analysis synthesizes 85 studies and introduces a unified mapping that links techniques to their efficiency and environmental implications. The findings show that although parameter-efficient methods, such as Low-Rank Adaptation (LoRA), adapters, and prompt tuning, effectively reduce adaptation costs during the adaptation stage, the overall computational burden remains high because of the reliance on large pre-trained models. Furthermore, although approximately 61% of studies consider computational efficiency, only approximately one-third provide an empirical evaluation of environmental impact. This highlights a clear imbalance between performance optimization and sustainability considerations. Our study emphasizes the need for integrated evaluation frameworks and energy-aware model designs. This contributes to a more structured qualitative understanding of the efficiency–sustainability trade-offs in T2I systems.
 
Keywords—text-to-image generation, diffusion models, zero-shot learning, parameter-efficient fine-tuning, computational efficiency, green Artificial Intelligence (AI)
 
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).