Users often encounter situations, where there is much background noise also recorded along with the human speech, that they are attempting to capture. This can be heavy ambient noise from near-by traffic, and A/C unit, wind, birds in a park, or others speaking close enough to the mic to be picked up. Most of the time, these extraneous noises are unwanted in the audio portion of the AV file.
Unfortunately, the best solution comes before one captures the footage. A good mic, good mic’ing technique, quieting the environment, or moving the taping to a quiet location are the best practices. Sometimes, many of these “cures” are not possible. Still, they are where one really needs to start.
When the audio has been captured, there can be major issues, when trying to remove the unwanted noise. There are some programs, and techniques that might be helpful at reducing these noises, but remember, these will only be so effective. If the noises are isolated to particular frequency ranges, outside of those of human speech, the prospect increases of success. A 60Hz hum is a good candidate for removal. Other human speech is certainly not. Wind noise can be broad spectrum can be very problematic, as if you remove it, for all overlapping frequencies with the human speech, those will all be removed, leaving the desired sound removed too. This is not what you want. One’s success will depend entirely on the frequencies of the undesirable noise, and how much of it, one can live with.
Now, there are some tools, that can be very useful, but only within the realm of the parameters discussed above. Regardless of which tools you use, you had better plan on spending a lot of time with a good set of headphones, and a lot of listening and testing.
The first tool that I turn to is Adobe Audition. It is a pro-level audio-editing program, and is not inexpensive. It is very powerful, but with that power comes both a learning curve and a lot of manipulation. One of its Restoration Effects is Noise Reduction (Process). This is used in conjunction with the Capture Noise Reduction Profile, which, when applied to a section of the offending noise, without the human speech, will map the audio profile of that noise. This profile will then be loaded into Restoration>Noise Reduction (Process). There, one has many controls for the application of the Effect, and the ability to Preview those settings. Plan on spending some time with this Effect, as it can be very easily overdone. When this happens, the human speech will take on a strong echo. As you reduce this undesirable effect, the ambient noise level will increase. Make settings, test - repeat - repeat - repeat. When you have balanced the noise reduction vs the distortion, that is as good as it gets.
From that point, one would then look to Effect>Parametric EQ (Equalizer) or Effec>Graphic EQ, to manually reduce certain frequencies (outside of the range of human speech), and also pump up certain frequencies that make up human speech - 125 - 250 Hz is about where most human speech occurs. These can be increased, but if your noise is full-spectrum, remember you will also increase any part of that noise, that resides in this range too.
After all has been done here, there are three Filters, that might help too: Low-Pass, High-Pass and the Notch/Gate Filters. Again, these can be very useful if there is zero overlap from the noise to the spectrum of human speech. Again, apply, test - repeat, etc.
One very useful Effect in Audition is Repair Transient. This can be used to remove pops, clicks and other transient noise. Many bird calls are Transients, and with a bit of work, can often be eliminated completely. I could not live without this Effect, as it can clean up so very much, leaving the noise floor untouched, but completely removing the Transients.
Now, buying Audition is not in everyone’s budget. One good freeware program for audio-editing is Audacity. It can work with many VST’s, and there are VST versions of several different EQ’s, plus High-Pass, Low-Pass and Notch/Gate Filters. I have never seen a noise-reduction VST that really worked, nor a Transient Removal VST, but they might exist. Note: while Audacity is free, only some available VST’s will be free. Some are more expensive than Audition.
Some years ago, I was working on a particularly badly degraded piece of audio. It came from a VHS tape of an old TV show, and apparently the tape deck had major problems, and maybe the signal was poor as well. This piece of audio just did not exist beyond this VHS tape, so I had to work on it. I spent a full day in Audition, and really was not doing well. It seemed that if I reduced part of the noise, some aspect of the audio disappeared also. Months before, I had purchased one of the Magix Music Studio programs. It came bundled with a little program called Audio Cleaning Lab. I installed both programs, but never touched ACL. It just languished on my Desktop. As I was about to give up on the restoration, I opened up all of my Magix programs, hoping to find a “magic bullet.” Nothing in the usual suspects, and then I opened up ACL for the first time. It had a simple little interface and looked more like a toy, than a serious audio program. There were default presets for several parameters, but not much more. Out of curiosity, I grabbed my WAV from Audition and ran it through ACL with the default presets. I was amazed at how much it improved on my day’s work. When the shock subsided, I decided to try it on the un-altered WAV. It was amazing. I then worked on the output from ACL, back in Audition, and got a useful file. That little “toy,” had done a fabulous job within the parameters and limitations of a program that does only one thing - cleans up audio. If one does a little additional work in a full-featured audio editing-program, like Audition, things can be amazingly good. Even then, one still must be ready to adjust, test, and repeat.
My audio “toolbox” runs from Audition to Audacity, and includes several of the Magix programs and many others. When trying to “clean up” existing audio, I am ready to use any/all of those, and be patient.
Still, recording the cleanest audio initially, will pay dividends. Cleaning up poor audio in post-production is limited and is very labor-intensive.
Another “fix” for bad audio is “dubbing.” This is the process of re-recording the audio in a quite location, with good mics, and the actors watching their performance on a monitor. This is the opposite of lip-syncing, but can be very effective, and is done in Hollywood often. Once a good clean dub has been recorded, it would be Imported into one’s NLE (Non Linear Editor) of choice and aligned with the Video portion of the Clip, that was used as a visual for the dub. The original Audio would be Deleted, or Muted. One of the nice benefits of doing things this way is that one can dub to an edited visual. If necessary, one could also do a tiny bit of Time Stretch on the dubbed Audio, to get perfect sync. This does get involved, but can be done. Many programs will allow one to alter time, but maintain pitch.
If necessary, other audio, besides just the dubbed human speech, can be added, where necessary. If one needs certain SFX (Sound Effects), to match visuals, these can be added. A second similar method is to do Foley Sound, where one finds items that will sound “like” what appears on screen. These range from things like “creating” thunder by shaking a sheet of aluminum, or similar. If one’s dubbing is also missing wanted sounds, using SFX Clips (either recording these, or finding them online) or doing Foley Sound can replace those audio elements that are now Muted, or Deleted.
I also highly recommend that one use a good pair of noise-canceling headphones, and apply a liberal dose of patience, when doing any audio work. Often, it can take far more time, than either setting it up correctly in the first place, or just doing a reshoot.
Hope that this helps someone,