How to use CoreML to accept multiple frames as input

I want to use CoreML to process video data. The ML model will take multiple frames as input. How should I get multi frames from ios and process it?

Thanks in advance for any suggestions.