I wonder if anyone has any advice for me? I had some audio knocked back for hum in the background . When I turned the volume up on the recording I could see what they meant. I’ve done a few room noise recordings, at different degrees of gain with my mic. I’ve recorded at different gains and the higher the gain, the more noticeable the hum. Probably self-evident. When I normalise it, it’s much of a muchness though. It’s noticeable enough to have been a problem so it needs to be addressed but I really don’t know where to start.
My set up: Apogee Mic, plugged into an iPad using Twisted Wave. I then dropbox it to my computer and manipulate it in Audacity.
On this occasion I put computer to sleep and turned off fish tank. I didn’t turn off the fridge or anything. For obvious reasons it would be nice not to have to do that, but it may be that that’s the noise.
I’m wondering if anyone has a minute to listen to my room tone recordings and give me some feedback? It may be that I just need to turn off more things. Or, can anyone suggest a tweak to my setup (probably not as it’s pretty minimal!). I need to get this nutted out because I really can’t audition until I get it sorted out. Which is a nuisance because a shiny new audition just appeared and I’ve got 24 hours to address it! LOL…
I have done nothing to these files, bar opening the wav files in Audacity, chopping out the bits and exporting it immediately as another wav. (No normalising or anything.) If I’ve done them properly you’ll hear a bit of drivel at the beginning (just me talking to get a baseline) and then the rest of the 10 seconds is just room noise. (I hesitate to call it room tone! It’s definitely noise.)
So I’m just wondering if anyone can point me in the right direction as I’m at the end of my, admittedly limited, technical ability. Is it ambient noise (in which case I can turn more stuff off), or is it something innate to my set up?
From listening to the first two samples, the problem is not actually “hum”, but just “noise”, particularly low frequency noise.
You can make quite a drastic improvement to that quite easily by filtering out the lowest frequencies.
The attached file has the settings for an Eq curve that is well suited for the job (it will be included in the next Audacity release, but for now it needs to be installed manually). low-rolloff-for-speech.xml (454 Bytes)
To install this Eq curve,
click on the “Save/Manage curves” button in the Equalization effect,
then click the “Import” button. Navigate to where you downloaded the “low-rolloff-for-speech.xml” file and select it,
then click the “Open” button.
You should now see “Low rolloff for speech” listed.
Click “OK” to close the curve manager.
To use the curve, first ensure that Equalization effect is open in the “Draw” mode.
Set “Length of filter” to maximum.
Click the “Select Curve” button and select “Low rolloff for speech”.
That should make a significant difference.
I would then suggest a little bit of Noise Reduction. Ensure that you are using Audacity 2.1.2 (look in “Help > About Audacity”).
Instructions for Noise Reduction are here: http://manual.audacityteam.org/man/noise_reduction.html
I’d recommend what we call “settings of the beast” Set all three controls to “6”.
(Don’t use the old “Noise Removal” effect for this, it is likely to cause more damage than good).
Regarding recording level, you could probably be a fraction closer to the mic and project a little. Avoid “talking to the mic” - imagine there’s someone across the other side of the table and talk to them. That will raise your voice level without turning it into “announcement mode”. The mic gain can then probably be about where you had it for the “Gain down a notch from high.wav” recording (that’s just a little low at the moment, but a bit closer and a bit more projection will fix that, AND it’ll improve the “signal to noise ratio”.
I am using 2.1.2 - I had a massive computer problem about a month ago and had to start from scratch with software (what a job!), so it’s whatever was most recent about 4-5 weeks ago.
I wasn’t being much of a voice actor - just talking so I could see the line moving. I was about the hang ten hand span away from the mic and speaking very slightly off-axis. I can certainly try a different distance from the mic and/or to try to speak a little louder.
The treble, yes, that’s been mentioned before. Noise reduction - this bit of audio was an audition for Voice Bunny, who required no noise reduction. So I wasn’t game, though I’d normally do a little.
I tried the low rolloff for speech, but I can still hear some noise. I pulled the length of filter all the way to the right - to 8191.
OK, so three more files, at various stages of processing!
Thank you so much for your help, much appreciated!
Guidelines for audiobook production also say “no noise reduction”. The important thing is that you should not be able to “hear” noise reduction.
Whether voice-over or audiobook, it’s all “art” and there’s no such thing as cheating in art If it works it works. If no-one can tell that you’ve used noise reduction, then that counts as “no noise reduction”. The bad thing is when you can hear that noise reduction has been used.
For audiobooks that’s a no no. The deathly black silence of deep space is quite disconcerting, and can be more intrusive than a “small” amount of hiss. Probably the same in voice over work.
If you need to eradicate all noise, then as Trebor wrote, you need to use a noise gate. However, if Voice Bunny say “no noise reduction”, then I’d guess they would dislike dead silence even more.
Within reason, the stronger the voice signal the better. Noise is all about relative levels. If your voice is only 10 times louder than the noise, then the noise will be very obvious. If your voice is 10,000 times louder than the noise, then the noise is insignificant. It’s the ratio of noise level to voice level that is important (called the “Signal to Noise Ratio” or “SNR”). If your voice is too loud, then you can record at a lower level, which will make the noise quieter.
The noise level in your new processed version sounds fine to me. I can’t hear that you’ve noise reduction, and I can only hear the noise because I’m listening fairly loud through headphones.
I’d also go for a slight increase in treble. About +2 dB from the Bass and Treble effect sounds nice to me.
Thank you for your help. I’ve been trying to upload some new files, with no luck. It looks as though it’s going to work this time!
Attached are three files. I’ve followed advice and got a bit further. The first file is as I recorded it; second one normalised only (ie, for the Voice Bunny audition) and the third is the best I could get it for a “real job”.
Oh gosh I just realised how I chopped off the words - really bad!!! I wanted to put a word in there but I thought I’d got a whole word. Oh dear.
Well, it’s certainly MUCH better than before. Whether or not it is “broadcast quality”? Don’t know. All the (pro) studio recordings I have have very little silence - they’re ads, so no long gaps for comparison. I guess all I can do is send off another VB audition from my home studio and hope for the best. I don’t want to do that until I’ve got it as good as I can though, because I’m not sure how many chances I’ll get and I don’t want to squander them.
The original one - that got knocked back - passed the ACX check plugin for Audacity, so as far as it was concerned, the noise floor was (just) below -60. Not sure how that squares with the noticeable noise. This version shows noise floor at -67 (with only normalisation); after I did the eq, a little noise reduction and then normalised, it was about -80. With my good headphones plugged into my computer, I could still hear some “noise” - it certainly wasn’t dead. All of which leads me to wonder if my concept of noise floor is right. More googling in my future! (Oh, no fiddling with the treble yet - though I agree - I just want to address one thing at a time.)
-80 db noise floor is very low. If they need it lower than that then it will need to be gated, but with the noise floor so low already, (a) I doubt that will be required. (b) if it is, then there will be no sudden “jump to silence” (which is the worst part of noise gating). Noise wise, I’d be happy with what you have.
Well, it was only -67 for the version without the noise reduction. But it was noticeably better than the -60 I submitted. I will do as requested and not do any noise reduction, but I’ll wait until the house is empty and turn off the power to the whole house to get it down as far as possible.
A noise-gate can chop off the the beginning & ends of words. The attack/decay time of the gate should be at least 100ms, and the threshold just a few dB above* the noise-floor. [* negative numbers , -50 is above -55 ].
That’s good to know, but I think in this case it was incompetence!!! I meant to chop to give just a word (for context and to give me something to “normalize”) and I really seem to have stuffed up! I was so focused on the silence at the end that I obviously forgot to go back and check. Not my finest hour.