Enter the keyword for the square and, if the puzzle gives one, the period.
Félix Delastelle, a French amateur cryptographer, devised bifid around 1901. It starts from a 5×5 Polybius square (I and J share a cell), usually scrambled with a keyword, and splits every letter into its two coordinates.
Decoding reverses it: turn the ciphertext into a line of digits, split the line in half (the first half is the rows, the second the columns), and read the pairs vertically.
For long messages, Delastelle split the text into blocks of a fixed length, the period, and applied the process to each block separately. If a puzzle says “period 5”, enter 5. With a period of 0 the tool treats the whole message as one block, which is how most short puzzles are set.
Each ciphertext letter is built from the coordinates of two different plaintext letters. That fractionation smears the letter frequencies, so the frequency analysis that breaks a substitution doesn't work directly. In a puzzle, the keyword is the thing to hunt for.
Each letter is looked up in a 5×5 Polybius square and its row and column are written beneath it. Read all the rows, then all the columns, as one long line of digits, and turn each consecutive pair back into a letter through the same square.
Long messages are usually split into blocks of a fixed length, the period, and each block is enciphered on its own. Leave it at 0 to treat the whole message as one block.
Every ciphertext letter mixes the coordinates of two different plaintext letters, so letter frequencies no longer map one to one. That mixing is called fractionation.