Now you just gave me exactly what I needed to prove that your encryption is weaker than using a proper one time pad.
Imagine the following algorithm:
For each possible string of characters (2^(8*n) possibilities, where n is the number of encrypted characters), apply your algorithm to convert this to a key. Then using this key, decrypt the text. If the decrypted text and the initial string of characters both were sensible in some way, we cracked the code.
Agreed that this would work? It's a slow algorithm, 2^(8*n) tries, and practically impossible. But not theoretically impossible.
I think you'll have to agree with this.
However, having a real one-time pad DOES NOT suffer from this flaw. It is theoretically, not just practically impossible to break. This is because you only have one string of characters to find out if it makes sense.
So, which part of this did you not agree with?