How Morse code represents text
Morse code encodes each character as a sequence of short and long signals, dots and dashes, chosen so that the most frequent letters in English are the shortest: E is a single dot and T a single dash, while rarer letters such as Q and Y take four symbols. The modern table is defined in ITU-R Recommendation M.1677-1, the version this tool uses, which is why it is called International Morse rather than the older American variant.
The code only works because of the gaps. Within a letter, symbols are separated by a short pause; between letters by a longer one; between words by a longer one still. Written down, those gaps become a space between letters and, in this tool, a slash between words. Without them, .... . could be HE or several other combinations, so a decoder fed Morse without spaces cannot recover the text.
Reading and writing the translator's output
In Text to Morse mode the input is upper-cased first, since Morse has no case, then each character is looked up in the table. Characters with no entry are dropped, and the complete list of them appears under the output as skipped characters, so a message with an é or a curly quote is flagged rather than silently mangled. Fix the text and the warning disappears.
In Morse to Text mode the decoder is forgiving about notation. Bullets and middle dots count as dots; em dashes, en dashes and underscores count as dashes; words may be separated by a slash, a vertical bar or a run of three or more spaces. A sequence that matches nothing in the table is replaced with a replacement-character glyph in the output and listed as unrecognised, so a stray minus sign or a prosign such as SK shows up immediately.
Audio playback and the speed setting
Play audio synthesises the current Morse with the browser's Web Audio API: a 650 Hz tone switched on for one unit per dot and three per dash, with one unit of silence between symbols, three between letters and seven between words. The unit length comes from the slider: 1.2 seconds divided by the words-per-minute value, the convention based on the word PARIS, which is exactly 50 units long.
At 5 WPM a unit is 240 milliseconds and each letter is easy to pick apart. At 40 WPM it is 30 milliseconds and the characters blur into rhythm, which is how experienced operators hear it. The timing ratios are fixed, so there is no Farnsworth mode (standard-speed characters with stretched gaps). In Morse to Text mode the player reads the Morse you pasted, which is a quick way to check a transcription by ear.
Limits of the table and when to use something else
The tool implements the ITU table and nothing beyond it. There are no accented letters, no prosigns such as SK or BT, no Cyrillic, Japanese or other national Morse variants, and no error signal. SOS is encoded as three separate letters with letter gaps, whereas the distress signal is properly sent as one continuous prosign. The three-letter version is what the audio plays.
Morse is an encoding, not a cipher: anyone with the table reads it. If the aim is to hide text rather than signal it, the ROT13 encoder does that job, and if you need text as ones and zeros for a lesson or a puzzle, the text to binary converter produces the bit pattern of each character instead.