separate audio by *beep*

I have a very large audio file composed of all the voice messages from my answering answer. Fortunately, my answering machine has a 3.5mm output jack, so I was able to record everything at a relatively high quality.

The problem I have now is that I can’t find any of my messages. They’re all together in one big audio file. At the end of every message, there is a clear and consistent beep sound. Is there any way for audacity to recognize this beep sound and cut out the segments between two beeps. In effect, this would create separate audio file for each individual message.

If audacity can’t do this, can any other program?

Find one of the beeps in the recording, select it by clicking and dragging the mouse across the waveform so that you select the beep and 1 or 2 seconds of audio either side, then use “File menu > Export Selection” and export the selected audio in WAV format. Attach the exported file to your reply so that we can take a look. (upload attachment tab below the forum message compose box).

here it is.

The message plays the message first. then says the date. and then comes the beep sound.

By selecting the beep and using Plot Spectrum you can see that the beep is at 1046 Hz.
Assuming that all of the beeps have exactly the same frequency, you can make the beeps easily visible with the following method:

  1. Make a duplicate of the recorded track (Select the track then Ctrl+D)
  2. Select the duplicate track
  3. Download the attached “1046.xml” file.
    1046.xml (187 Bytes)
  4. Open the Equalization effect from the Effect menu and select the “Draw Curves” mode
  5. Import the 1046.xml file. This will load a new preset called “1046”
  6. Apply the “1046” equalization preset to the duplicate track and it will virtually silence everything in the track except for the beeps, which will become significantly bigger and should be very easy to see.

Well yes but did you listen to what that curve did to the messages? :wink:

I was able to label each message on its own excluding the beep by some experimentation and only moderately distorting the sound.

  • Apply heaviest Leveller setting three times
  • Draw a not too extreme EQ curve like this which makes the tone quieter rather than louder then apply three times
    EQ curve.png
  • Truncate Silence to 20 milliseconds so there is little gap between the words and between words and tone
  • Analyze > Sound Finder with silence level of -20 dB and minimum silence of 0.18 s

Export Multiple will now export each labelled message on its own.

If the message with the effects applied is repeated, it looks like this:
labelled messages.png
Of course the other messages would need to have similar audio characteristics to achieve only one label per message, but you could conceivably go through making one label for any multiple labels that a message created.


hi guys - thanks a lot for your responses! I had sort of forgotten about this project until my answering machine filled up again…

I appreciate your solutions to make the beeps more visible, but I literally have many hundreds of messages. Each one is separated by that “beep” sound. Manually selecting each message would still be an arduous task. Is there a way - a script of some sort maybe? - to automate this?

You could try using the technique described in my earlier reply ( to exaggerate the beeps on a duplicate copy of the track, then use Silence Finder ( to add a label before each beep, then use Export Multiple ( on the original track using the labels that you have created.

Hi Gale - that solution worked well, but a bit too well. Every time there was a bit of silence in the voice messages, audacity would mark it as a new message. Is there a way to mark the beginning of a new message based specifically on the audio characteristics of the “beep”?

No, my method made the beeps quieter so that Sound Finder would ignore them and label the messages as the sound, but Steve’s method makes the beeps louder so that Silence Finder thinks the messages are “silence” and so labels them.

But using Steve’s method you can try increasing the silence level in Silence Finder so it doesn’t think the noise in the recording is silence.