keras tensor mainpulation output to next input

[JohnyBeGood]

Hi,
Since 2 week, I try to merge a dense layer output to a LSTM input to create a kind of auto-feed ML (today forcast is add to tomorrow input)

the code for the moment is:

main_inputs = Input(shape=(30, 6))
lstm1_out = LSTM(100, return_sequences=False)(main_inputs)
output1= Dense(1)(lstm1_out)

[b]output1_mod = something to modify an fit output to nest LSTM input.... [/b]

merge_input = K.concatenate([main_inputs, output1_mod],2)
lstm2_out = LSTM(100, return_sequences=False)(merge_input)
finaloutput= Dense(1)(lstm2_out)
...

here a picture of what I try to do...



Thank so much for your attention...