Separate LSTM networks with different sequence length

I am quite new to this, and still trying to figure things out. Any help is appreciated.
I am trying to create two separate lstm networks, which would take in different sequence lengths. So, for example lstm1 would take in sequence length of 6 and lstm2 would take sequence length of 12. Afterwards, the outputs of both networks should be put together to produce an output of sequence length 1.

Can this be done?
I have read about pack_padded sequence… But not sure if it applies to hear.