Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{176969,
        author = {Nikhil Chandurkar and Prof. Supriya Jawale and Prof. Kaveri Deosarkar and Prajwal Kothare and Aditya Meshram and Gokul Chaudhary and Harsh Khandelwal},
        title = {A Comparative Review of Fine-Tuning Techniques and the Evolution of Image Generation Models},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {12},
        pages = {804-811},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=176969},
        abstract = {This paper explores advancements in image generation, focusing on four fine-tuning techniques: DreamBooth, Hypernetworks, Textual Inversion, and Low-Rank Adaptations (LoRA). DreamBooth personalizes image generation by associating a unique textual identifier with the subject’s visual features, while Hypernetworks generate dynamic weights for adaptable model behavior. Textual Inversion introduces new concepts through compact embedding vectors, and LoRA adapts models efficiently with low-rank updates. The evolution of image generation models from Stable Diffusion to Flux is examined, highlighting theoretical and institutional shifts. Empirical comparisons show DreamBooth excels in output quality but requires substantial computational resources, whereas LoRA offers significant efficiency and lower computational costs. Practical implications of these techniques are discussed, with DreamBooth preferred for high-fidelity applications and LoRA for resource-constrained environments. Advanced LoRA implementations, such as rank and alpha parameter optimization, precision format comparison, and multi-LoRA composition, are explored. Future research directions include cross-architecture compatibility, hardwareaware fine-tuning, and ethical considerations, providing valuable insights into the current landscape and future directions of finetuning techniques in image generation.},
        keywords = {Image Generation, Fine-Tuning, DreamBooth, Hypernetworks, Textual Inversion, LoRA, Stable Diffusion, Flux Architecture},
        month = {May},
        }