Take the 2-minute tour ×
Sound Design Stack Exchange is a question and answer site for sound engineers, producers, editors, and enthusiasts. It's 100% free, no registration required.

I have an orchestral recording with an audience. In one part of the recording, there are shock chords (loud, short chords played by the whole ensemble) at 0 dB with about two seconds of 'silence' between each. Obviously the silent parts aren't actually silent: they have the usual venue and audience noise, as well as the sound of the shock chord dying away and echoing, so they don't get any quieter than -72 dB.

Some child in the audience found the whole thing quite amusing, and giggles in one of the silences. It's less than -48 dB, but really damages the effect. It's during the quietest part of the silence, and well away from the following shock chord, so it should be easy to remove, but it's giving me trouble. How can I remove this unwanted sound?

I've tried using frequency-based filters, but the giggle is quite broad-spectrum: it's mainly around 50 Hz, right over the background sound, so I lose too much that way.

I've tried running Audacity's "noise removal" effect on the passage, priming it with the background from another silence, set to "isolate" instead of "remove" noise, but even at 48 dB it leaves way too much giggle.

I've also tried pasting a section of one of the adjacent silences, making sure it's the same length and getting the zero-crossings right, but because the echo is slightly different for each chord, it leads to a jarring transition. Each shock chord is the same chord (though the balance between instruments is slightly different), so it seems like it should be possible to make this technique work. Maybe I just need to do something clever with cross-fading to a second track that's a copy of the adjacent silence? How do I make that work in Audacity?

share|improve this question

migrated from avp.stackexchange.com Jan 27 at 14:59

This question came from our site for engineers, producers, editors, and enthusiasts spanning the fields of video, and media creation.

    
The best tool for something like this is any program with a spectral editor, or an editor that lets you make selections by frequency and time rather than just time. I would use Adobe Audition for such a task. I don't believe Audacity is capable of this. That sounds like an easy fix, if you can't get it to work I can probably look at it for you. –  Michael Nov 22 '13 at 21:44

2 Answers 2

Your instinct is correct - a reasonable way to deal with this using Audacity would be to copy the section to a second (stereo) track, shifted in position so that an earlier piece aligns with the giggle.

Then use the 'envelope' tool to shape the volume profile of each track. Since the replacement section was earlier in time, lower its volume enough to match the track it's covering, fading the offending track down as you fade the replacement up, and vice versa.

share|improve this answer
    
It's very difficult to get a seamless crossfade with the envelope tool. You have to align everything by ear –  Dan Hulme Nov 19 '13 at 16:55
    
Indeed, you have to refine almost everything by ear. (-: The alternative is to apply fade in, fade out and volume change effects per track, which IMO is more work for no net benefit. But by all means use what's comfortable. –  Jim Mack Nov 19 '13 at 19:16

The technique I used in the end is quite close to Jim's, but I thought I should make a new answer to set out the steps one-by-one. There are three important parts to the process, to avoid having to line things up by eye or ear, and to make the paste seamless. They are:

  • Using the Cross Fade In/Out effects to give a constant-loudness fade
  • Using Region Save/Restore to select the same time period in each track
  • Using Move Cursorto Selection Start/End to select a period exactly adjoining the previous selection

The process is as follows:

  1. Copy a nearby, clean 'silence'. In my case, I made sure to include the peak of the shock chord too, to make it easier to line up.

  2. TracksAdd NewStereo track.

  3. Make sure the new track has the right sample format and rate, and then paste into it, with the cursor about the right point.

  4. Drag the pasted track so that it's under the noisy 'silence', and line up the peaks.

  5. In my case, the offending sound wasn't loud enough to show up in linear view, so at this point I switched both tracks to Waveform (dB) (in the little menu on the left). This view makes quiet signals more visible.

  6. On the pasted track, select the start of the track: from the start of the pasted signal (the shock chord) nearly up to the offending sound, but leaving a gap in between. ctrl+L to silence that period.

  7. EditMove cursorto Selection End. Then use shift and arrows and/or shift-click to select from there to the start of the offending sound.

  8. EffectPluginsCross Fade In.

  9. Now you need to select the same region in the original track. To do this, use EditRegion Save, then click on the original track, then EditRegion Restore. (It's called Selection Save/Restore in older versions of Audacity.)

  10. EffectPluginsCross Fade Out. Now your original track fades out while the new one fades in, in exactly the same region.

  11. Now to select the offending sound, using EditMove cursorto Selection End followed by shift and arrows or shift-click. Make sure to leave another margin after the sound for the corresponding cross-fade.

  12. ctrl+L to silence that annoying giggle.

  13. Now repeat steps 11 to 8 to select the margin you left after the offending sound, to cross-fade the original track in, and the new track out.

If you screwed up and didn't manage to entirely cut out the giggle, or you cut off the next shock chord or pasted halfway through a cough, you'll have to undo several operations and redo them by hand. When you've listened and are satisfied that you've improved the recording, TracksMix and Render will get you back to one track.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.