Talk:Sanitizing speech recordings made with portable audio recorders

From Audacity Wiki
Jump to: navigation, search

I guess as a newbie to audio editing in general and Audacity specifically, a breakdown of what the different Effects 'do' to the file, and how they 'typically' change what I (and my audience) hear is needed.

I have learned from my editing using Photoshop, that there are ways to shoot yourself in the foot and then try to put on your bulletproof shoes. So I will be putting down my best guesses, and hoping others will test and improve or put down what they do (even if this is based on other editors).

Someone must have written this down somewhere.

Not finding them yet, here is some initial experimentation on a recording that was probably made with a camcorder built in microphone in a larger meeting room. There was a ringing type noise:

1st, heavy use of the preview button is suggested. 2nd, save your work often as you go in case your improvements are not improving the overall sound.

Glame Bandpass Analog Filter - seems to strip out high end hiss. I tried the preview a number of times and went with a setting of Center Frequency of 1200, and Bandwidth setting of 3800hz.

GLAME Butterworth Lowpass - seems to strip out more high end hiss. I tried the preview a number of times and went with a setting of Cutoff Frequency of 3000 hz, and a Resonance of 1.21 - This make the ringing worse, so I used the Edit|Undo.

Hi DamianPodCaster

If you want to colour the text, I recommend using

<font color ="red"> text to be coloured </font>

because enclosing text in double square brackets creates a link to another page on the same Wiki.

Some documentation on the LADSPA plug-ins you are using is here , if you have not already seen it. The built-in Audacity effects are described on the Effects  page of the online manual .

You can also try Noise Removal or Notch Filter (a more precise type of filter). Noise Removal works much better in Audacity Beta 1.3.3 or 1.3.4 than in 1.2.5/6.

The effect settings which will work best will never be the same for different types of noise.

Saving frequently is a good idea in 1.2.5/6 anyway, but it does not of itself let you get back to a previous project state unless you do File > Save Project As.. and save the Project to a new name e.g. with the effect settings used in the file name to save confusion. Doing this does take a lot of disc space, and Edit > Undo is preferred. You can undo any change stepwise while the project is open. This means that if you make a good change (1), a bad change (2) then a good change (3), you cannot get back to change (1) without undoing the second good change (3).

Beta 1.3.3 or 1.3.4 also lets you use much longer preview times if desired.


Suppressing spikes and noise

It is a very important article you are maintaining here! Audacity seems to be a fantastic tool but it is not very easy to know where to start. I started using Audacity only two days ago so I have not much experience with it yet, also have no previous practical experience of digital sound processing. I only have some basic understanding of sound recording technology and a concept of what could characterize a good voice recording. From the perspective of an "interested amateur" on this, I'll make an addition in the article about ideas on noise reduction and spike removal.

Using stereo could open up other possibilities perhaps

Here is a idea, maybe it is crazy. In the tutorial section I saw instructions how to remove vocals from a stereo recording, effectively making a "karaoke" track. Well, if it is possible to remove information that is common to both channels (e.g, the voice of a singer panned in the middle of the sound stage), then it should be possible to enhance this common information instead while suppressing everything else. Maybe this could be done by summing the two channels of the "karaoke-ized" stereo track into a mono track, then summing the two tracks of the original stereo recording as well making it another mono track. If those two mono tracks are summed, maybe after first having inverted one of the tracks, everything but the voice should be suppressed.

Has anyone tried something like this? It sounds like it has a small chance to work in theory at least... Otherwise I feel quite tempted to try, just have to find a good stereo microphone first.

In principle you can write algorithms to isolate material common to both stereo channels in one step - all the third-party Windows plug-ins mentioned on Vocal Removal Plug-ins have this theoretical ability. It doesn't help a chair scrape recorded into both channels, obviously.

It seems that moving from mono to stereo recording opens up fantastic opportunities with tools like this! Very interesting. It is probably worthwhile to see what other have done in programming special filters for Audacity, maybe there already exist good solutions.

The problems addressed by this article/talk exist not only in the "audio player-recorder" area but also on the audio portion of video camcorder (both analog and digital) recordings.