Film Post Production

Hi all,

Just wandering if anyone has got any tips for cleaning dialogue recorded on location. The problems usually consist of wind, air conditioning fans and plane rumble. Massive thanks to anyone wanting to share their magic.



very interesting question for me too…
my experience is: what is bad recorded stays bad. Therefore the set recordist has to be a pain in the neck for the whole film crew at the location shot till all the unwanted noise is cut out or doing ADRs with the actors afterwards… But U can experiment on how high you can set the low cut on voices, female voices sometimes you can cut up to 140hz. you can try an expander (is there a lv2 plugin which is usable as an expander??) to separate the dialog more from the background noise.
My experience with noise reduction is very little, but I think especially for the human voice where the ear is trained to hear microscopic changes in pitch and colour noise reduction is often useless. But I havent tried expensive noise reduction software yet…

Short version is, Linux is the wrong platform for noise reduction at this point.

Longer version. Take a look at the “Wires for Empathy” trailer I posted not long ago. All those SFX were recorded by me, most with a cheap Tascam handheld unit and some with rather extreme gain. And believe me, almost all of them went through some noise removal. Or take a look at some of the pieces I work on for my work. You can get rather acceptable results from modestly priced software on OS X or Windows these days. I use a combination or two tools right now for my work, Izotope RX and WaveARTS Master Restoration Suite. I find they both have their advantages and disadvantages, with Izotope typically being used first, and WaveARTs if needed to clean up after Izotope. I also can run WaveARTs on Linux with Festige, and could do Izotope as well if I could get the authorization dialog to generate a machine code, but that is a different story.

The thing to keep in mind is that you need to know where to draw the line. There is a point where you will just destroy to much of the sound to be useful. It is better to be more conservative than not. And the funny thing is, you may find that you remove some sounds only to put in similar sounds artifically afterwards, and it MAY be an improvement depending on what you are doing. Perfect example…

One of the few ones of these pieces where I didn’t use any music to mask some of the artifacts heard, but instead put in seperate ambient noises in the background, the end result was I was able to lower the overall noise level in the recording a few dB compared to the end result, even with adding in more ambient sounds later.

On the native Linux front, I tried tools like Audacity, and while it has improved recently, it still doesn’t compare to either tool above to be honest sadly. I never had much luck with GWC, and the impression I have gotten is that development on it hasn’t progressed much in years.

I really need to write up a blog post about this…


seablade, very helpful comments for me too, thanks,
yes masking with music or other ambient sounds sometimes make things better!
another tiny question on your example video: did you record the dialog with a boom mic (which?) or lavaliers? I really like the sound quality.

is there a way to use the linux dsp 500 series as an expander? or maybe the calf compressor? that would be very helpful for me , like I usually dont want to remove background noise but maybe lower it in volume just a little bit…


That one was a boom mic IIRC, a M/S combo of an AKG c451b and CK something, whatever their Fig-8 capsule is in their blue line series. If you look around there is an example of one recorded with a couple lavs in a fishing boat, but sadly one of the voices especially didn’t get a good position set for the lav so it was to far up and the voice was pretty muffled as a result and the voice quality isn’t great.

In general for this place I am working with a combination of static booms (Audix Microbooms generally as they are what I have), a boom pole (The above mentioned combo), or lavs (Coutnryman B3s through a ULX wireless system).

I don’t believe LinuxDSP has an expander no. The Calf Gate has an expansion mode IIRC, but I wasn’t to impressed with it when I tried it, but it can be a good starting tool I suspect. And yes some slight downwards expansion can work wonders if applied judiciously.


Thanks for the input seaplade, always good to know how other guys do their jobs. I ll try the calf plugin again and will see if i can get a suitable effect…

is there a way to use the linux dsp 500 series as an expander? or maybe the calf compressor? that would be very helpful for me , like I usually dont want to remove background noise but maybe lower it in volume just a little bit...
There isn't an expander in the 500-series at the moment, except for the gate, which is a kind of fixed (high) ratio downward expander.


Yea no offense intended but I would never use the gate you have for a downwards expander, the ratio is far to high, but that wasn’t the intent either I am sure. That being said I obviously do use a lot of your plugins on a regular basis for other things:)


…but would be a great feature having a expander option in the gate plugin in the future? :slight_smile:

@Seablade looking forward to your blog :wink:

someone gave me a hint to these lv2 plugins: theres one that works as an expander… anyone has some experience with them? I tried the expander on some dialogue but didnt get useful results, but maybe that was my own fault…

When I eventually get some time (unfortunately I’ve had to devote more resources than anticipated to other plugin updates for non-linux OS etc as a necessary part of ensuring some income for the company) I hope to look at how to model a dynamics / expander / gate processor based on a very well known pro-audio console design which I have some knowledge of.


Not that this should come as a surprise, but you make a decent true expander(Meaning adjustable ratio with timing, not just depth control) and I would probably be first in line to pick it up:)


Hadn’t seen those before, will have to try them out sometime.


@ linuxdsp that would be a fantastic surprise :slight_smile: So I m just starting to forget about it again :wink:

well configured a gate like calf gate or even the ladspa gate would help, however i’ve used Mixbus’s Dynomite on a drums bounce track that had a good deal of reverb and it managed to expand it and at the same time of course shortening the reverb, giving it much more dynamic range, much cleaner sound… however that worked on a song that has several other instruments, very far from a dialog case.

I don’t know if that would help this type of cases, but i’d give it a try…

@ fernesto thanks for the hint. dynomite is a native mixbus plugin? in the essentials bundle? havent heard of it before, i ll check that…

@calimerox: yes it is native mixbus, as far as i know it is already included, no adittional plugins pack required, its quite minimalist only one parameter, check it out on youtube.

Have you guys tried gate? It’s very flexible!

It is, however, not a downwards expander sadly. It certainly could possibly be useful, but what I truly want is a good downwards expander really.