The first 10(2n-1) bits, where n is the number of unique characters in the uncompressed file, contain a preorder traversal of the Huffman tree used to encode the file.
The next 2 bits are always 11, which can be used as a method to verify file integrity.
The next 8 - n bits of data contain a padding to make the data structured as bytes again.
The next 64 bits of data contain the number of bytes that the uncompressed data takes up.
The data section contains all the compressed data, Huffman encoded. The file is padded with 0s to contain modulo 8 bits.