Text-to-image generation has seen remarkable advancements in recent years, particularly with the rise of diffusion models, which have significantly improved the quality and realism of generated images. These models have demonstrated substantial success in multilingual text generation, including English, Chinese, and Arabic, etc. However, the generation of text in less widely studied languages, such as Khmer, remains a significant challenge. Khmer, the official language of Cambodia, presents unique linguistic and visual characteristics that make it difficult to generate accurate and legible text within images. This paper provides two primary contributions. First, we establish one of the earliest baselines for Khmer visual text generation using diffusion models in real-world visual scenarios, offering a foundation for future studies on low-resource script synthesis. Second, we conduct a detailed evaluation of common rendering issues specific to the Khmer script and suggest practical solutions to improve model performance. Also, we publish the experiment data, including text prompts and generated results of all models’ evaluation for all group categories at Khmer-Benchmark-VisualTextGeneration( https://gitlab.univ-lr.fr/ksaly01/khmer-benchmark-visualtextgeneration .