Some help would be much appreciated.
Was out on the weekend and had a mini interview with someone and i used my old Canon GL2 with its standard mic, basically the background noise is quite loud car sounds and its sort of cut out the voice all together making it very hard to hear!!!
Is there anyway i can fix this background sound and bring the voice out stronger so its not a total waste of the interview, help would be appreciated thank you!
For my new camera i will be buying a normal microphone to plug in so i dont have this issue again.
+ Reply to Thread
Results 1 to 17 of 17
Your problem is one that has been encountered by a great many film makers.
There are two ways to overcome this problem.
First, you can try to edit your sound in a program like Audition or Cubase. However, this technique is extremely difficult, to a level where even the most qualified and skilled sound editing geniouses might not succeed. The problem with this method is that machines do not percieve sound like human beings and probably cannot differentiate voices from bgm, thus, if the "background" noise is so loud to make the "foreground" voice unnoticable, they lose their positions and become one. Separating them in this situation is, as I mentioned before, mostly impossible.
The second method you could try, and this is the one that 99% of film makers choose after recording bad audio, is never to record an interview in a place where you can hear cars and buses.
This is a basic rule in filmography, and I am afraid that the only way to get decent audio is to record good audio in the first place. Otherwise, anything you try to alter your bad recordings will sound horrible.
In other words, if it possible, redo the interview. That is your only real option.
Failure is a greater step towards perfection than Success. With Success we are satisfied. From Failure we learn.
Or, if the Shame of Failure is great enough, we commit Seppuku, and reach perfection in Heaven. Remember this as long as you are alive, young Shuriken.
*Video Ninja fades into the darkness*
With most audio editors you can use a low pass and a high pass filter and cut off higher frequency sounds and lower frequency sounds. That may help a bit with a voice recording. Try dropping frequencies below about 200Hz and frequencies above about 2500Hz. May help a little. I use Audacity for that most times. Noise reduction doesn't help much unless the interference is steady, like 60Hz hum. If you play with the high and low frequency settings, you might be able to improve the audio a little. Audacity lets you 'revert' so you won't make any permanent changes till you save the output.
I would expect a improvement at best. not a fix.
And welcome to our forums.
Along the same theory as redwudz's suggestions, audition/soundbooth have a spectral view where you can select visually specific frequencies. (either cut them out or reduce them etc...)
So if the car sounds exist in a different range / pattern than the human speech you might be able to distinguish them and edit it out
If the car noises are constant background noises and continuous, not random noise intervals, you can capture a footprint with it at an interval where the speaker is not speaking, and digitally subtract it; this is the "usual" method of noise reduction with audio editors
Many editors also have "enhance speech" type plugins which act to bring out the freqencies of normal human speech
As mentioned all of these methods usually will only slightly improve, not fix.
poisondeathray and redwudz, while your techniques might make help distinguish the speaker, they will also make him sound pretty horrible. When cutting frequencies you always lose some of the original quality and can never compensate for that!
A true Video Ninja must plan his moves in advance and execute them perfectly. A room must be designated for the interview, or at least a quiet place. If possible, a microphone should be used in proximity to the speaker, so that background noises might get easier to clean in post production.
Sound is just like video, Shuriken! If the source sample is of bad quality, there is only that much that can be done about it. No post-production can replace good footage or good sound. Alterations will give you results that will sound unnatural. Only the strongest and most talented sound editors can walk this fine line and reach the other end. I suggest you do not try it before training with a high level sound master.
Sound is 50% of the movie - Steven Spielberg
*Video Ninja fades into the darkness once again*
You're completely right videoninja! We were just making some suggestions that may help to salvage what WeAreDrift has.
I would have offered my time machine, but it's in the shop right now
A time machine could be a very useful tool for a ninja.
I should look into this.
*bow of appreciation*
Your other option, which is also widely used in the film industry, is to shoot on location, then re-record the dialogue in the studio and mix it into the audio afterwards (usually rebuilding all the audio at the same time and using none of the location material). Yes, it is slower and an imposition on your talent, but if you can't record the audio on location correctly, this is the only option for getting good quality audio after the fact. Attempting to salvage very noisy location recordings is often a fool's errand.
videoninja, you joined just today, have made 9 posts as I write this, some of them very helpful to people, but I find your style of writing very annoying. Is there any chance you can write in a more 'normal' fashion?
Or am I the only one bugged by this?
Originally Posted by manonoI think,therefore i am a hamster.
Ask me in a month if it's still entertaining...(if he/she's still here in a month).
back on topic, I concur with what has been offered. If your "background" is more like a middle- or fore-ground, you're wasting your time. The fixes that can be done might help when you have a clear dichotomy between fore-and back-ground. Even then, they will lose something of the fore-ground's clarity.
Best simple device is what used to be known as a Burwen Noise Supressor. Automatically narrows the BP when no guide track (vocal range should be this guide) present, expands to wider BP when guide track is loud. Used the masking effect to minimize the background noise (but also adds sometimes noticeable "pumping").
If the Canon GL2 records in stereo (I don't know if it does) you may be able to eliminate some of the noise by phase shifting one of the channels. Noise will arrive at both mic cartridges at about the same time, while the voice may be closer tro one than the other. Shift the phase of one channel 180 degrees and the sound that is common to both channels will cancel out. Of course, if the soundtrack is mono this won't work and if the speaker's audio is 'dead center' in the stereo image it won't work either.
To expound upon what olyteddy just said, you might be able to squeak out a little bit better sound by manipulating the stereo channels...
It's unlikely that the sound environment for this recording would have been a directly equi-distant vocal subject as well as auto traffic directly in-line behind that subject. So tweak the relative phase and/or delay of the channels...
If the subject is dead center and the traffic is all around or off to one side, sum the channels equally (or close to equal--see what works best).
If the interviewer is dead center behind the camera, subtract (difference) the channels equally (but only during the time when the interviewer talks). Crossfade between these.
If the subject is off center, use a phase-shift effect (look it up), or delay the channel that the subject is tending toward. Adjust the delay along with the relative volumes when blending (adding/summing) the 2 channels to find the sweet spot where the sinewave onsets are in- or near- sync for the subject. Of course, they'll be out-of-sync for the traffic, so it'll get a minor filtering or dimming.
Absolute phase reversal is probably unlikely to be needed, but you can always add it to your toolkit.
There are a few other tools that might help with the isolation problem (signal channel multiplication / double cross-ring modulation with sideband pre and post filtering / guide track vocoder filtering of the main signal), but they are totally experimental (I could talk about it but I'd have to kill you) and also produce varied results and take an OBSCENE amount of time to produce any results.
Hope that helps,