Morse Audio Translator

This page generates Morse sound, it does not receive it

Choose text input for a known phrase or Morse input for already separated groups. The page converts the input, builds its timeline and plays tones only after you press Play. It does not upload or decode a recording and does not use a microphone. If your aim is to find the words in a recorded signal, a manually prepared transcription can be checked here, but the listening and boundary identification remain yours.

Inspect warnings before listening. Unsupported characters in text mode are reported and omitted; the remaining valid output can play. In Morse mode an unknown group stays visible as [?] in the transcript and prevents a valid timeline. A reassuring sequence of beeps can therefore represent an incomplete text input, and a readable partial transcript is not enough for full playback. Check the exact source, not only whether sound is available.

The character, signal and duration displays measure different aspects of input. A signal is an individual dot or dash. Duration is the calculated span from the first mark to the last, excluding a final word gap and repeat pauses. It is not a recording of speaker latency, screen delay or successful reception. Very short timings are rounded in the display, so use the worked examples below for understanding ratios rather than treating the visible decimal as a precision measurement.

The four kinds of duration that make code readable

In ordinary spacing, the dot defines one unit of time. A dash lasts three units; marks inside a character are separated by one unit of silence, letters by three and words by seven. ITU-R M.1677-1 specifies these ratios. The dot and dash use the same selected pitch. Their duration and order carry the information; a higher note is not used to mean dash.

At 20 WPM the dot unit is 1.2 / 20 = 0.06 s. A dash lasts 0.18 s and a one-unit inner gap 0.06 s. With spacing speed also 20, a letter gap is 0.18 s and a word gap 0.42 s. A is a 0.06 s dot, a 0.06 s pause and a 0.18 s dash: 0.30 s in all. N reverses the mark order but has the same total duration. Listen to both before assuming equal duration means equal character.

The difference between EE and E E is silence, not marks. At 20/20, EE has one dot, a three-unit letter gap and another dot, totalling 0.30 s. E E substitutes a seven-unit word gap and totals 0.54 s. There is no extra letter gap added on either side of that word gap. This comparison is a useful diagnostic when another application or your manual transcription seems to use unusually long spaces.

Character speed and Farnsworth speed solve different problems

Character speed ranges from 5 to 40 WPM and controls marks and the one-unit gaps between them within a character. Farnsworth spacing ranges from 5 WPM to the current character speed and controls the longer gaps between characters and words. The default is 18 WPM characters with 14 WPM spacing. If the character rhythm is clear but the next letter arrives before you can identify it, reducing spacing speed provides more time without stretching the character itself.

The WPM reference is the 50-unit word PARIS, including its following word gap. It contains 31 units of marks and inner gaps, plus 19 spacing units: four letter gaps of three units and a final word gap of seven. With Farnsworth timing, the 31 units stay at character speed and the 19 are stretched to achieve the selected reference rate. ARRL’s Farnsworth timing calculation gives the derivation; this is a calibration convention, not a claim that every five-letter word takes the same time.

For character speed c and spacing speed f, this engine uses dot = 1.2/c seconds and spacing unit = (60/f - 31 × dot)/19, never less than a dot. At 20/10, the dot remains 0.06 s but the spacing unit becomes about 0.218 s. The letter gap is then about 0.654 s and the word gap 1.525 s. Lowering spacing is not a general slowdown of every element: A still lasts 0.30 s.

PARIS by itself ends at the last mark here, so at 20/20 its displayed duration is 2.58 s, not 3 s. The missing 0.42 s is the final word gap used in WPM calibration. Do not divide an arbitrary clip's word count by its duration and expect the control value to match. Digits, punctuation and unusually short or long character patterns all change a real message's duration.

Control tone, level and repeated listening separately

Pitch ranges from 300 to 1000 Hz and volume from 0 to 100. Choose a tone that is comfortable on your actual speakers or headphones; there is no universally best pitch implied by the default 640 Hz. Changing pitch does not change code, speed or gaps. Page volume is separate from system and device volume, and zero is silent even if playback appears to be running.

Repeat starts another copy of the same signal after a pause of at least half a second or the selected word gap, whichever is longer. That pause is not part of the duration display and is not exported. For a comparison, first hear one pass with repeat off, then use repeat to focus on an uncertain character. Stop before changing text or settings; the page also stops automatically on those changes to avoid overlapping output.

A practical listening sequence is three or four known characters, a short word using them, then an unfamiliar combination. Keep one setting fixed while changing another. If you always replay HELLO, the repeated L is useful initially, but you may learn the phrase's melody. A second word helps you test whether L is recognizable outside that sequence. The chart provides an isolated reference, while quiz and practice supply other prompts; none guarantees a particular training outcome.

Use WAV for one portable signal, not the whole browser session

The WAV download uses the same selected character and Farnsworth timing as playback, with the chosen pitch and volume. It contains mono, 44.1 kHz, 16-bit PCM audio for one signal. It does not include repeat, light, vibration, an answer narration or the original text as a spoken introduction. Export at the final settings and test the saved file, rather than assuming a previous preview proves that the new file sounds right.

Both live playback and WAV generation reject signals longer than 90 seconds. This is calculated duration, not a character-count allowance. Lower Farnsworth speeds, digits and punctuation can make a short-looking message unexpectedly long. Split a lesson into labelled excerpts at sensible word boundaries. Removing gaps to fit the limit would change what the reader or listener has to interpret, not merely compress a recording.

Keep the text answer and both speeds beside the WAV in your own lesson notes. If a player changes playback speed, it changes timing and may also change pitch depending on that player; the original WPM settings then no longer describe what the class hears. Compare other generators using their reference conventions and trailing silence as well as their labels. Agreement on the marks alone does not establish equal spacing.

Troubleshoot silence, interruptions and misleading comparisons

If there is no sound, start with the simple checks: nonzero page volume, unmuted tab and device, selected output device and a deliberate Play press. Try a short valid E at comfortable volume before investigating a long clue. Browser audio support is required; a blocked or unavailable audio API cannot be repaired by changing Morse punctuation. A generated WAV can be tested separately in a media player to distinguish page playback from the rest of the audio path.

If playback stops, check whether you edited the message, changed a control, selected a new chart entry or hid the tab. Those actions cancel the previous playback. There is no resume-at-position feature; restart the signal deliberately. Long signals can fail the duration limit, and invalid written groups can prevent playback altogether. Read the notice and the unknown-group warning instead of repeatedly raising volume.

Optional light follows a screen indicator, not a camera torch. Vibration depends on browser and device support. Neither is embedded in WAV, and their timing can lag audio. Flashing may be unsuitable for some people, so use it only deliberately and respect reduced-motion settings. For a timing lesson, the diagram and written examples remain useful without motion. No combination of these options establishes a reliable closed-tab alarm or emergency signalling system.

Why did operators learn a rhythm rather than a picture?

The Smithsonian's telegraph sounder describes a receiver whose electromagnets moved a lever to produce audible clicks. This is not the smooth sine tone generated by the page, but it shows why listening became a practical way to receive marks. A chart tells you a character's pattern; a timed signal makes you distinguish its duration and boundary before the next character arrives.

The Farnsworth approach addresses that gap between knowing a printed pattern and recognizing it quickly. The ARRL paper explains keeping a character's shape while allowing longer pauses for recognition and writing. Its historical training standard used an 18 WPM character rate, but the method itself is not tied to that number. The controls here let you compare other supported rates without claiming that a certain setting is a tested curriculum for every learner.

Maritime stories add context when used carefully. The 1909 Congressional Record on Jack Binns records Binns's CQD calls from Republic in 1909, and the Science Museum's account of Titanic wireless discusses the 1912 distress traffic and attentive reception aboard Carpathia. Sound mattered because people were listening to messages with context. Generated SOS is useful for hearing short and long marks, but it is not an authentic recording of either event and should not be labelled as one.

Frequently asked questions