UniBioPAN is a versatile peptide classification model inspired by video processing. On one hand, we considered that peptide sequences can be viewed as a sequential string with temporal characteristics. On the other hand, molecular structure graphs have been widely employed in drug development due to their inclusion of structural information such as atoms, the connections between atoms, bond angles and etc. Our methodology treated each amino acid as one frame within a time sequence. We employed a CNN to extract features from each frame, employed a BiLSTM to capture the temporal relationships upstream and downstream in the sequence. UniBioPAN was applied to fit 31 different bioactive peptides (BPs) and compared it with the current universal method, UniDL4BioPe. When applied to a baseline dataset of 37 different BPs, UniBioPAN demonstrated superior classification performance, interpretability and broad adaptability. The results indicated that our model was suitable for various BP classification tasks, offering significant potential guidance for the development of BP classification models.