Image captioning efficiently bridges the disparity in visual content and linguistic expression, providing an automated solution that transforms images into descriptive text. This capability is invaluable in various domains, from enhancing accessibility for visually impaired users to enabling more effective image search and retrieval systems. Automating caption generation simplifies tasks that traditionally require human input, making it a powerful tool for managing and interpreting large volumes of visual data. This paper presents an image captioning system utilizing deep learning, combining convolutional neural networks (CNNs) utilized for extracting visual features and recurrent neural networks (RNNs) employed for generating coherent descriptions. The proposed model is tested on datasets on MSCOCO 2014 primarily and FLICKR 30k, with detailed analysis of hyper-parameter effects on word accuracy which increased from 0.534 to 0.743 during training, while validation accuracy improved from 0.509 to 0.709 by the 20th epoch. Evaluation metrics BLEU-1 to BLEU-4 scores of 0.59, 0.45, 0.37, and 0.30, respectively. Perplexity dropped from 11,528.86 to 6.47, system performance, CPU/GPU usage and memory efficiency showing the model’s growing confidence. These results confirm the ability to generate relevant captions. The study also highlights challenges in balancing model complexity with performance, suggesting future improvements in computational efficiency and caption detail and having real-world applications include e-commerce, marketing, and education.