Morse Audio Translator

1. Get the audio in. Drop a recording into the box (WAV, MP3, M4A, OGG and friends all work, depending on your browser's audio support), or follow the microphone button to let it listen live. A built-in demo sample loads into the tool with one click if you want to see the pipeline work before hunting down a recording.

2. Press translate. Nothing moves until the button is pressed; then a translation counter runs smoothly from 0 to 100 while the sweep crosses the range where Morse tones live, locking onto the strongest steady pitch it finds and tracing exactly when that pitch turns on and off. You will see the detected tone and speed in the stats line, so you can tell at a glance what it decided to listen to.

3. Read the result. The English text lands in the box with a copy button and a play button that speaks the translation back in Morse. Anything the tool could not match confidently comes out as a question mark, so you know exactly where to listen again, and if the recording was too rough to read a single character, the box says so plainly instead of showing symbols.

Recordings vary a lot from clip to clip: the tone, the volume, the background noise, and the timing are all different, and a real sender rarely keeps textbook spacing. This tool treats the recording as imperfect and estimates tone, noise, and rhythm before translating anything.

The tone comes first. A scan across roughly 300 to 1300 Hz finds the strongest steady whistle in your audio, wherever your app, radio, or keyer happens to put it. Then the tool learns what silence sounds like in this specific recording, hiss, room tone, static and all, and judges every moment of sound against that measured floor rather than a fixed volume. Because the floor is measured from the recording itself, a quiet phone recording and a noisy radio clip are judged by the same standard.

The rhythm is where recordings really differ, so the timing is never assumed. Every burst of tone gets collected and their lengths are clustered into short and long groups: the dots and the dashes, wherever your sending put them. The same clustering applied to the silences separates the tiny pauses inside a letter from the pauses between letters, and those from the long pauses between words. The same measured process reads a machine key with textbook timing and a nervous beginner whose dashes run long and whose gaps wander, though very irregular sending can still blur some characters.

Last comes honesty about confidence. A group that matches a known character cleanly gives you a letter. A group that does not gives you a question mark in the text, because a wrong confident answer is worse than an honest one.

• Trim the dead weight. Cut long stretches of silence or music at the start and end of the file so the tool spends its attention on the keyed part.
• Keep the tone alone. Music, speech, and overlapping voices sit right on top of Morse frequencies and smear the rhythm. A clean sidetone recording is usually easier to translate than audio mixed with music or speech.
• Favor a steady pitch. Practice apps and keyers hold one tone, which the translator loves. Whistled Morse works too if you can hold the note reasonably level.
• For microphone capture, get close. A phone a hand's width from the speaker in a quiet room beats a laptop across the desk, and the live level bar shows you exactly how strong the pickup is.
• Mind the speed range. Roughly 5 to 35 WPM is the comfortable zone; beyond that, dots become too short for real-world recordings to survive noise.
• Use the stats as feedback. The tone and speed the tool reports are its best guess at what it heard, so if the tone looks wrong you know to fix the recording rather than blame the translation.

It is at its best with practice-app exports, WAV and MP3 recordings of a single sender, hand-sent keying with a normal fist, and radio audio where the tone stands clear of the static. In noisy conditions the stats line drops from strong to fair to weak so you know how much to trust the text.

Hard limits, so you can plan: uploads are capped at 15 MB per file, and recordings longer than five minutes are turned away, because the whole file must fit in the browser's memory at once. Trim the recording to the keyed part first; a shorter clip translates faster and just as well.

It cannot perform miracles. Two stations sending over each other produce tangled rhythms no single-tone listener can untangle. A sung or heavily bent pitch that drifts hundreds of hertz loses the lock. Speeds past about 35 WPM squeeze dots shorter than room noise allows. And audio where music or voices drown the tone gives the tool nothing to hold on to. In all of those cases you will see question marks marking the parts the tool could not read confidently, so it is worth checking the text against the original audio. A cleaner recording often improves the result, though some signals can remain difficult.

When you have text rather than audio, the Morse code translator on our homepage converts it in both directions, the NATO alphabet page covers spoken spelling-out, and the Learn Morse Code page trains the ear that makes all of this automatic.