We present a BLIP-based multimodal system for image-based humor generation submitted to SemEval-2026 Task 1 (MWAHAHA), focusing on Task B1. Our approach fine-tunes a vision–language model on meme-style captions and handles animated GIFs via representative frame extraction to generate culturally grounded humorous captions.