FMA-Net is a deep learning model specialized in video super-resolution and deblurring. It is designed to restore videos of low resolution and blur into high resolution and clarity. The model achieves this through a combination of flow-guided dynamic filtering and iterative feature refinement using multi-attention techniques, which are effective in handling large motions within the video. This results in a joint super-resolution and deblurring of videos. The model boasts its simplicity in structure and notable effectiveness, making it suitable for wide application in video enhancement and editing fields.