Pixnatte: A BLIP-Based Image Caption Generation System for Context-Aware Visual Description

The exponential expansion of digital media has created a pressing need for automated, high-fidelity visual description systems capable of functioning at scale. Existing image captioning approaches frequently exhibit trade-offs between generative creativity and factual accuracy, limiting their utility in real-world applications. This paper presents Pixnatte, an image caption generation and visual question answering system built upon the Bootstrapping Language-Image Pretraining (BLIP) architecture. The system integrates BLIP-based image captioning with BLIP-VQA to enable contextually coherent, content-specific descriptions of arbitrary input images. Pixnatte is trained and evaluated on established open-source datasets including MS-COCO, Flickr30k, and Conceptual Captions. Experimental results demonstrate that the proposed model achieves improved descriptiveness and reduced hallucination rates compared to conventional deep learning baselines. The system shows strong generalization to unseen images and supports potential deployment across accessibility, digital media, and multilingual communication domains.