This uses near-optimal encoding.

The tree is encoded in-order, as in the leaves are sorted left to right. It is encoded in code-length, ascii value form, with code-length and ascii value both being bytes.

After that it's traditional huffman encoding.

Some space could be saved by removing the first bit from the ascii values since ascii is only 7 bytes and doing the same for the length since the tree cannot get longer than that for the same reason.

From source, make sure you have cargo installed and run

cargo run -- {file_you_want_encoded} -o {output_file_name}