Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment - View it on GitHub