I am starting to work with Ardour for video (commercial, documentary, and journalism) and there are two key details that are important to me. I would like for users to share their techniques in problem solving. I am using Ardour on GNU/Linux, 64 bit.
First, noise removal. For example, if I am interviewing a shop owner while several refrigerators generate mains hum. (the owner cannot shut off the refrigerator or leave the shop) Or if I have a street interview and there is the sound of traffic in the background, or even just wind. Proprietary software has tools to minimize background noise, but maybe there are Linux-friendly plugins that do the same.
Second, to optimize the voice of the interviewee, such as running audio footage through normalization, compression, and the like.
To get good sounding recordings you must do many things correctly at the recording stage. Errors made there may be difficult and often impossible to correct after the fact. And always it will take time and effort. Sometimes you can make a bad recording sound better with eq, compressor, noise reduction etc, but you can never make good sound out of a bad recording. Things to consider while making an interview:
Choose the least noisy environment. If you record video at the same time, don’t forget to listen to the sounds of the environment. If you shoot a person talking, then the most important information is not in the picture, but in the speech. Choose the best environment for sound not the picture.
If the environment is noisy, then go as far as you can from the noise source.
You want the speech signal to be much louder than the noise in your recording, so move the microphone as close to the speaker as possible.
Use a low sensitivity dynamic microphone, a lavalier or a directional shotgun microphone in a noisy environment.
When outside always use a wind shield on the microphone. Wind noise might be impossible to get out of the recording or at least it will force your eq settings so that you will have no say in how the speech sounds.
With a directional shotgun microphone position the speaker so that the noise comes from the side to the microphone, as sound coming from the side will be attenuated the most.
If there is a noise coming from a certain direction, try to make all recordings so that you keep the same distance and direction from the noise source, otherwise the noise floor will jump up and down when you cut different takes together. After the interview you may want to record the environment alone, because you may need to add in more of the environment noise in your mix if the noise floor jumps from one take to another.
If your recording device has the option to cut down lower frequencies while recording, do this while recording speech. The lower frequencies will mostly contain noise and frequencies you don’t want to have in the first place (wind noise).
Always listen to the sound with a good pair or headphones while recording. If you do not do this then you will miss things like wind noise, the sounds you hand holding the microphone makes or a nearby gsm telephone whos radio frequency transmission makes loud sound in the recording.
At the post processing phase cut down lower frequencies from speech as the bass frequencies mostly will make the sound muddy. How much you should cut bass depends on the case.
Compress the speech or drive it up against a limiter. This will make the speech better cut through music you may have in the mix.
@Hartzel, thanks for your feedback. I do have a pretty good lav mic and a Zoom mic for now. I am starting out so I cannot afford a shotgun mic just yet. Recording circumstances so far have not always been ideal and mistakes happen so I try to clean up the best I can through software. Given the landscape of software out there, which plugins do you use for Ardour? Also, what OS do you use? For example, I know that Ubuntustudio has a broad range of plugins for Ardour. Given that there are interviewees in different environs and I want to get ideal sound for video, which plugins would you use?
I’m sorry I can’t be of more help here. I’m a hobby musician and mostly record electronic instuments, so there is no noise problems there. At work we do tv audio post production and have found the iZotope RX tools to be very useful in “sound restoration”. We use an older version of the package, I think the separate “restoration” tools are now packaged in a product called Advanced Audio editor. We found the tool “Dialog Denoiser” to work quite well, but I’m not sure what’s the name of the product where it is included now. Also at the moment we are experimenting with Zynaptic Unveil for removing unwanted room echo from interviews and it seems quite a promising product.
But enough with that commercial stuff, you wanted open source tools I would suggest as Edward Diehl did earlier in this thread that you try the Noise Removal tool that is built in Audacity. Here is a video of how it is used:
Basically using the Noise Removal tool is a 2 step process.
First you need to tell the tool what you consider to be unwanted noise in the recording. This is done by selecting a short area of the recording where there is only unwatented noise and nothing else. Then you open the Noise Removal Tool and click on the button “Get Noise Profile”. The tool now inspects the selected area and what frequencies appear there.
Then you select the whole recording and open Noise Removal tool again. This time you select how many decibels you want the noise to be attenuated. Be careful with this setting since removing too much noise quickly brings in some “hollow” sounding artefacts in the audio. Start with a low value like 6 dB’s, and if that is not enough undo and try a higher setting. Noise removal makes you choose between the original noise and the sometimes quite nasty sounding artefacts that the noise removal process creates. You need to balance between those
A shotgun mic might not help in a noisy environment, especially if you are recording indoors, since the noise tends to bounce off the walls and around the room. When you turn the mic away from the noise source, you might get it again reflecting of a wall
At the moment I use Gentoo Linux, before that I used UbuntuStudio happily for a couple of years. I don’t think that you need any fancy plugins for the stuff that you do. In tv audio post production the main tools we use are eq, compressor and limiter. These will get you most of the way there. It might help to put each speaker on his own audio track, so that you can eq and compress each person separately. Then you probably need to record some volume automation for each audio track to level out volume jumps and make each speaker as loud as the others. Sometimes you need some noise removal (that might be done with audacity) and sometimes some de-essings for taming those sharp s - sounds.
I’m just beginning to explore the plugins that are available for Ardour, but at the moment I use EQ10Q as my eq (or it’s smaller version EQ4Q), SC4 compressor from Steve Harris, Invada Early Reflection Reverb (for adding room echo in music), ZamExcite to create some more high frequencies when needed (for musical instruments).
The demo at http://isse.sourceforge.net/index.html (linked above by Edward Diehl) looked pretty impressive to me. The first example was easy, but the later ones were really quite sophisticated for free software. You can’t do good noise removal with a simple plugin though; it 's a question of analysis first, then experimental adjustment for best compromise as described in other posts above.
I’d re-iterate MHartzel’s earlier advice about mic use and choice. The absolute best way to keep noise out of any recording is not to record it in the first place. A lav mic, useful though it may be for convenience and hands-free operation, is no substitute for a hand held mic right in front of the speaker’s face. Consider an EV 635A (a classic ENG mic that’s been used for decades for this job) for those really difficult interview environments. You can hold it 2cm from the speaker’s mouth and because it’s omnidirectional there’s no proximity effect, and anyway it doesn’t have much low end, which is good because you really don’t need it for this application. It’s a lot cheaper than a shotgun, and more foolproof!
Just an update here. Upon reflecting on mhartzel and anahata’s comments, the point here is that no matter what you do, sometimes undesirable noise seeps into a recording. You do everything but whether it is mains hum, some wind, seepage from outside, stuff happens. You record the best you can but you still gotta clean up. Now, the Windows/Mac world has LOTS of tools to handle this. Someone pointed out Izotope, which is great. But the question remains, what about the Linux world? We still pretty much have Audacity and some gating/eq stuff from LV2 plugins. However, breaking this is the most compelling candidate so far: https://github.com/lucianodato/noise-repellent One company also told me they are making something for the Linux world as well, but it is not ready yet. Furthermore, a program called Cinelerra also has noise reduction tools, but it does not seem to be pro-level yet. And that is the key, professional-grade noise reduction in GNU/Linux et al. It is happening, just slowly.
Noise Repellent is not at the level of iZotope Spectral Denoiser but is pretty clean for -10 dB SNR. It should work as expected for the type of noisy environments you mention. Be aware that it can’t eliminate the noise completely but for 5 to 10 dB of reduction is most of the time ok. Make sure to read the README file as it contains recommendations for a cleaner reduction too.
Hello Lucianodato, thanks for making this! My research indicates that I do not want too much noise reduction as it ‘deadens’ audio, but something to take out background hiss, traffic hums, passing cars, AC units helps. When do you suppose your plugin will be ready for the repos? (such as KX Studio).
The last time I tried Soundsoap, which was admittedly a while back, Izotope handily beat it. THe advantage to Soundsoap was that it was simpler to operate, but with minimal time invested learning Rx, you got much better results.
That being said, for the ‘cadillac’ of noise removal, most people would point to CEDAR, or another piece of software I can’t think of right now and is now going to bug me until I can.
Algorithmix, that was who I was thinking of. Their Renovator software. I think most people would put that and CEDAR as the top these days, but I could be wrong, I haven’t used them myself so can’t speak to it either way, but saw some impressive demos of it a few years back.
I’ve seen Cedar Demos and tried noise reduction on their demo computer. It is good but extremely overpriced. You can get 90% of the same results by doing it with iZotope RX4.
If you are working for a national security agency and need to clean recordings done secretly, then cost does not matter and you can buy Cedar. I think that is much of Cedars business anyway. But for average workshop RX4 is at a good price / performance point. Haven’t tried Algoritmix.