crypteria 7

Critique design
pass->hash 1024 bytes
text->1024 bit chunks ->
each 9 bits from start of hash, generate fill order in chunk
i.e. each bit from 1024 bitchunk is given new place [0-1023],each next bit is fiven new place[0-1022],[0-1021]... from next 9 bits in hash, next chunk fill order starts with 1 bit right to get 9bits slightly to end(different 9 bits).
XOR the result with bytes from end of hash
[Optional step for sending secure data: each byte is diffused into a bitstring with each bit given N bit random data gap(gap length calculated from current 9 bits from hash),similar to previous crypteria ciphers]

Assuming good faith (which is a bit of a stretch in your case), nowadays we already have plenty of fairly secure symmetric cryptographic algorithms, and most of them are pretty fast too. Yet another "fairly secure" algorithm is of therefore no use. You should rather look into designing "P=NP"-resistant (by assuming that SAT can be solved in O(n²)) algorithms and provable security.