#image-to-text

X-Fusion introduces a dual-tower architecture that adds vision capabilities to frozen large language models, preserving their language skills while improving multimodal performance in image understanding and generation.