There's a lot of background noise with a high pitched whine in this interview I'm trying to work on but I cannot seem to get rid of it with Waves C4 or EQ. They recorded the interview with the Zoom H4n on-board mics..Do you guys or gals have any advice?

Here's a sample: http://soundcloud.com/mitchell-scott/interview

7 Answers 7


I agree with Bruce's post. Izotope is the plug-in I'd reach for first...but yeah, the signal:noise ratio is pretty low, so there'll be a fine line between improvement and turning the audio into aquatic swimmyness. The fan noise is also pretty high-pitched, so I'd worry about EQing into the women's voices. If the audio was otherwise pristine I'd also suggest trying DeVerb to get that room 'verb outta there, but I think the background noise is the right battle to choose.

If you don't have Izotope, it's a bit spendy. Is this audio that will sync to video? If not, it might be worth a reshoot, cost/time-wise. You might also want to do an EQ pass to bulk up the voices as well, which feel a little thin in the low-mids. But I can also nearly guarantee that anyone who's purchased Izotope has never regretted it. :-)

  • When I read your answer I read "DeVerb" as "D-Verb" and I was terrified for a second and thought "Nooooooo!!! Nathan, Nooooooooooooooooo!!! Please tell me you don't use D-Verb!!!", but then re-read it and I'm fine now. :-)
    – Utopia
    May 31, 2011 at 6:12
  • (not to say that D-Verb doesn't have it's uses. On a snare, the Non-Lin alg can sound great. :-P)
    – Utopia
    May 31, 2011 at 6:12
  • @Utopia I used to turn my nose up at D-Verb but, IMHO, it's great for simple room verbs. And so, so DSP efficient! May 31, 2011 at 13:03
  • @Roger Yes, yes. I know. I just like making people laugh :-)
    – Utopia
    May 31, 2011 at 17:27

You are going to have a hard time with that audio.

As the interviewee's are quite off mic and lots of room noise, a lot of what you do will impact on their vocals. You can use a notch eq filter to try minimise the fan noise in the background.

The other option is to use a noise reduction plugin where you set the noise sample and then apply it to the clip. If used correctly, you will get a bit of noise reduction (you cannot push it with that audio you have) and hopefully bring out the vocals more.

Other guys on here might have some more effective ways at cleaning this up, but the old saying is "You cannot polish a turd", but you can sure try.

Good luck.


If this is what you have to work with, I personally think it's better to not try to drastically reduce the noise and leave it somewhat untouched and smooth out the transitions and edits to make it sound natural in and out of words and sentences. Even with top CEDAR gear, I don't think you'll end up much better off. The key to something like this is SLIGHT and GENTLE noise reduction - otherwise like NoiseJockey mentioned, you'll be swimming with the sea monkeys in no time.

I see this a lot in work my peers do - they take a very noisy recording and try to suck all of the noise out of it and they invariably cut right into the voice and make it sound worse than without the noise reduction. "But the hiss is gone!!" Yeah, so is the audience's understanding.

Just my .02.

P.S. Don't forget to give the recordist who did this a hard time and help him do a better job next time. And if it was you, learn from it and make sure you get a better recording next time. If I were to receive this quality of recording for mixing I'd instantly reject it as unmixable and I'm sure a lot of other studios would do the same.

  • @Utopia Hahah heck no I was not the recordist. I think it was the guy operating the camera. I don't know any recordists that would just record an interview with the Zoom on-board mics. I still want to make the voice intelligible to the audience, and cancel out the noise as much as possible like Bluesman69 did in his example. May 31, 2011 at 18:03
  • +! I LOVE the CEDAR. Would love to enter into an affair with one if only it wasn't so damaging on the wallet. Quite an amazing tool. The C4 when used correctly and gently finessed across a scene can take out quite a bit of noise, but the CEDAR DNS is on a whole other level completely. I agree, some residual noise is best left in to keep the dialogue alive. I like ttry and find that perfect balance between intelligible dialogue, but not sterile or artifact-ed. Feb 26, 2012 at 8:22

I would go for the Izotope denoiser, it should be able to bring down the background noise by about 12 dB with just a few artifacts, if you use C or D mode.

Render the audio with audiosuite and then put it on a track with a C4 to see if you can expand it a little bit afterwards.

Use a bit of EQ, you can perhaps tone down the noise with a broad dip with a low Q.

Just be careful not to do too much harm to the the dialogue while removing the noise.


Hi Mitchell Does it need to be pristine or just for the dialogue to be clearly audible? The noise signature is well defined and easy to separate, so you should be able to remove most of it using any decent NR plug-in. The dialogue will definitely be more audible, but the process may introduce a pumping effect and some artifacting. You can always add a bit of white noise or a soundbed afterwards to mask this.

I hope you don't mind, but I ran your sample through a basic iZotope Rx Nr algorithm and posted the result here: http://snd.sc/llUiI2

This was pretty quick and dirty, but you should be able to achieve better results doing multiple passes in small increments.

  • @Bluesman69 The dialogue only has to be clear. I actually just got Izotope RX2 not too long ago, but haven't really gotten the hang of spectral repair yet. Is that what you used to get rid of the noise? May 31, 2011 at 17:00
  • @Mitchell Scott Sorry, I should have clarified. I used the DeNoise module, as this is most effective for broadband NR. Your recording contains sections where there is noise but no dialogue, which suits this type of NR process. I'll try to add a quick walk-through as soon as I get a chance. The learning curve is not that steep, so you should be able to get usable results quite quickly. Spectral Repair is the most powerful component of RX, but is most effective in reducing or removing impulse noises such as coughs, banging doors, cell phones, etc.
    – Bluesman69
    May 31, 2011 at 22:03
  • @Bluemans69 Thanks a lot man. I appreciate it! Jun 1, 2011 at 2:31

Hi Mitchell

Here is the quick walk-through I promised:

  1. Select a few seconds of noise
  2. Open DeNoise and click Learn
  3. Click the menu arrow next to Presets, select Add Preset, type in a name and close DeNoise
  4. Select a few seconds of wanted dialogue, re-open DeNoise and select your preset
  5. Select the Simple tab - this should be fine for your source material
  6. Select Algorithm D - provides best results 99% of the time
  7. Set NR to 10, Smoothing to 7, click Preview and listen
  8. If you hear artifacts, set the smoothing up a bit and Preview again
  9. Select Output Noise Only and Preview to ensure you are not losing dialogue
  10. DEselect Output Noise Only and close DeNoise (NB Don't forget this step!)
  11. Select the whole file, open DeNoise and click Process

Repeat the above process until you are happy with the results.

This is pretty much the standard NR process for broadband noise, although I have biased it a bit to suit your source material. The Advanced tab gives you a lot more control over your noise profile in terms of the type of noise you are targeting and the manner in which artifacts are handled. You can usually obtain better results, but it takes substantially longer.

The noise in your recording is mainly broadband, but there appears to be a tonal element to it as well, so you would probably link them in the Advanced tab anyway. The real secret to successful NR is multiple incremental passes using smaller (8-10dB) values. Yes, higher values will mean more noise reduction, but will also introduce noticeable artifacts. Also, just in case it is not clear, you must either overwrite your existing preset or add a new one for EVERY subsequent pass.

Your source material will also benefit from higher Global and Fine Smoothing in order to minimise any possible pumping effect from the noise reduction, so once again the benefits of dealing with them separately will be negligible.

Finally, there is often a tendency to want to remove too much noise. Only remove what you have to and no more, otherwise your recording will start to sound artificial. Sometimes I even add in a bit of white noise afterwards to make it sound more natural.

Hope this helps!

  • @Bluesman69 Thanks so much man! I'm going to give it a whirl tomorrow. Jun 2, 2011 at 6:41
  • Thanks these are some helpful tips, but just wondering is this any good NRing? it was done using Cockos' Reaper free plugin pack and audacity's. If your wondering no, I didn't record this my dad loved that bird in the mornings while we were in India and he recorded them with my H4n. soundcloud.com/stephensaldanha/nr-test Jun 5, 2011 at 22:29
  • @Stephen There is artifacting but, as the sample is only 128kb/s, it is hard to say how much is from NR. I do think you're trying to take out too much noise and are removing wanted material as well. It is important to remember that some of the "noise" you are removing is actually natural ambience that belongs there. I would suggest just pushing it a bit to the background, not removing it entirely. Try 2-3 passes using smaller NR increments, around 6-8dB, rather than 1 pass. Also, I noticed some handling noise, for which spectral repair is most effective, although this can be reduced manually.
    – Bluesman69
    Jun 6, 2011 at 10:45

Maybe this programm wil do the trick: http://www.algorithmix.com/en/renovator.htm You can also find a movie about it on soundworkscollection.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.