Hi there,
While recording some podcasts I managed to succesfully reduce the background noise to really low levels. The inter speech interval is virtually a silence now.
The noise removal settings I used are:
Noise Reduction: 35
Sensitivity: 0.6
Frequency Smoothing: 150
Attack Decay Time: 0
My problem is that only during speech there is a kind of hiss together with the voice.
If somebody can give a name, a link, an explanation… any insight would be nice.
Thanks!
Ops, there was a reply here but it’s gone!
Anyway, Trebor, I tried different combinations of Noise Reduction and Sensitivity but they don’t seem to provide me with a good result.
I am uploading three small files, 2-3 seconds each.
I’d appreciate if you could listen to them and give me some advice.
The name of the files express their Noise Reduction / Sensitivity levels.
For example:
R10 - S10 Stands for Noise Reduction = 10 and Sensitivity = 10
Thanks!
Try amplifying the audio a bit first, then run this code in the “Nyquist Prompt” effect first:
(highpass8 (notch2 s 120 10) 100)
Then use Noise Removal with something like:
Noise Reduction: 8
Sensitivity: 12
Frequency Smoothing: 200
Attack/decay: 0
Steve’s noise gate will squelch the noise between words, e.g. …
But noise gates will not reduce the noise during speech.
Noise reduction isn’t perfect. That’s why professionals still record with good equipment in soundproof studios.
The best thing is to record with a strong analog signal (A strong voice, fairly close to the mic) for the best-possible signal to noise ratio.
Or depending on what you are using, you may need to upgrade your hardware. If you are using your computer’s microphone input, they are notorious. Other options are a good-quality USB mic (the kind of mic designed for music or voiceover), or a good mic and an external USB/firewire interface, or a good mic into a mixer with the mixer’s line-out going onto your computer’s line-in.
Noise Reduction only works well during human speech if you surgically turn it off during actual words. If you don’t do that, human speech turns into underwater aliens. You can set it to try by reducing smoothing and attack to zero. Then it will try to reduce vocal noise.
I know you’re asking yourself “what good is it?” Not much. See: Do a good recording in the first place.
Koz
Steve:
I’ve applied the code you gave me and performed the suggested Noise removal.
I’m uploading both files now, original and modified. See what you think. The hiss during the speech is definitely better but the overall noise is still too loud. I tried changing to other combinations of Noise Reduction and Sensitivity… had nothing better than this one I just sent.
DVDdoug:
Perhaps my recording apparatus is not good indeed, but it’s not so bad. I’m doing something wrong for sure, this is what I have:
Mic: Samson C03U USB.
Other Related stuff: Pop Filter, Shock Mount, Mic Windscreen.
Studio: I’m in China right now, so I managed to cover an entire room with studio foam for next to nothing.
I’m pretty much sure my quality bottleneck is my computer and related hardware, what do you think?
Just another thing I have to ask, if this “speech hiss” is something that I can’t do anything about it in the after processing. Is there some trick during recording time that would prevent this hiss of happening or perhaps decrease its intensity?
Thanks！
That’s probably about as good as you’ll get with that source material.
You could use the Noise Gate (as suggested by Trebor) to make the “silent” parts more silent (or even totally silent) but it will not reduce the noise during the talking.
The single biggest improvement that you can make is to record at a higher level.
Technical babble:
Currently the peak level is -16.8 dB. Unless the person talking suddenly starts shouting you can safely increase the recording level by 10 dB. The best way to do that is to get the microphone closer to the person speaking, and if possible get them to talk a little louder. The noise level in the original recording is -44.4 dB and the noise level after processing is -61.2 dB, so the “dynamic range” between peak signal level and peak noise level is 27.6 dB before processing, 44.4 dB after processing. If you can get the recorded signal (voice) level higher by 10 dB in the original recording, then the dynamic range between peak signal and peak noise will improve to at least 54.4 dB. I say “at least” because Noise Removal is more effective with low level noise, so the better the signal to noise ratio is in the original recording, the more effective Noise Removal will be.
Practical stuff:
If you get the voice a bit louder in the original recording it can make a big difference to the quality of the recording.
In Audacity 1.3.13 you can grab the recording meter with the mouse and pull it out from the main Audacity interface, then stretch it to the full screen width. This will make it a lot easier to see your recording levels - aim for a peak level of about -6 dB.
Steve,
Got it!
I told my staff to stay as near to the mic as possible but since the material they are supposed to read is too long, they ended up laying down on the chair while reading it, thus getting far from the mic.
I’ll pay more attention to the recording levels next time.
I believe my sound card/OS doesn’t allow me to change the input level directly from audacity interface. But i can do it in the Control Panel.
Thanks.
It’s the sound into the microphone that you need to increase.
With USB devices, increasing the volume in software (in the Control Panel) usually does the same thing as using the “Amplify” effect in Audacity - the signal level will increase, but the noise level will also increase by the same amount.
Give them upright chairs, or better still, high stools to sit on 
Haha, will do.
Should get “pop shield” … http://en.wikipedia.org/wiki/Pop_filter
or make one … http://www.hometracked.com/2006/05/28/homemade-pop-filters/
I have already, I listed before the equipment I was using.
Perhaps my recording apparatus is not good indeed, but it’s not so bad. I’m doing something wrong for sure, this is what I have:
Mic: Samson C03U USB.
Other Related stuff: Pop Filter, Shock Mount, Mic Windscreen.
Studio: I’m in China right now, so I managed to cover an entire room with studio foam for next to nothing.
I’m pretty much sure my quality bottleneck is my computer and related hardware, what do you think?Just another thing I have to ask, if this “speech hiss” is something that I can’t do anything about it in the after processing. Is there some trick during recording time that would prevent this hiss of happening or perhaps decrease its intensity?
Thanks！
But it is weird, even with the popfilter and windscreen I can still hear some air being blown. Chinese has some very aspirated syllables.
If it’s sibilant “sh” “ch” “zh” sounds they can be attenuated digitally with a de-esser VST plugin, (there is a free one called “spitfish”).
Before-after example attached, (I may have overdone the de-essing as I don’t speak Chinese, the effect is adjustable).
Live recording is always a juggling act between overload and noise. Overload and clipping (popping) is immediately fatal to a show, so you don’t want to do that. There is one other non-technical point missed. Is it always the same speaker that pops and has sibilant problems? Once you get beyond Skype quality recording, the possibilities for problems becomes vast. I know people that can’t speak into microphones. They sound terrible in any but the most top quality, protected microphone. You generally don’t want those people announcing for you.
“Background Noise” is the other extreme. People pay many dollars to get a microphone pre-amplifier that makes the voice louder without adding hiss. That’s in addition to the high quality of the microphone itself. Past a certain cost, the two come separate.
USB microphones kill you because the critical amplifiers and microphone processing are all built in. There are no options. If the particular quality doesn’t work for you and doesn’t fit your circumstances or presenter/announcers, sell the microphone or put it in the garage.
Post production patching, as you’re finding out, rarely works well and in any event, it doesn’t take many podcasts before you get so tired of messing with the effects that you run out and get a better microphone that doesn’t need all that.
Once you decide to perform theatrical editing and effects, you fall into the 10 to1 rule. It takes 10 hours to edit a one hour show. People always poo-poo that rule until they start actually editing and are realistic about counting up all the hours. How long have you spent on this so far?
Another fuzzy generality: You use the tools for the one show your grandmother performed just before she died. There is zero possibility for a reshoot and the show has to work. The other fuzzy generality is by the time you realize you need most of these tools, it’s too late to do a good job.
Koz
Nice Trebor, I didn’t know such a thing existed. I’ll take a look.
kozikowski
Is it always the same speaker that pops and has sibilant problems?
Usually male, with the girls there’s no problem.
If the particular quality doesn’t work for you and doesn’t fit your circumstances…
I think it does, I gathered some practical tips in the last couple of days and I think the quality now has improved a lot.
But if I do decide to chance my mic, what would be a natural upgrade? Just remember, if I decided to buy the USB mic in the first place it’s because I’m not a professional audio engineer/musician and I was looing for something simple to use and resoanable priced.
How long have you spent on this so far?
I don’t have any fancy effect inside the podcasts (they are lesson transcripts actually), I think I’ve been spending some like 5 minutes for each recorded minute. But it depends a lot on the speaker, some people don’t make mistakes at all, others keep coughing while recording, making the after editing a pain.
What do you think?
Thanks！
I think Audacity 1.3 has the ability to add labels on the fly. When the presenter makes a fluff, add a label to make it easier to find later.
USB microphones were designed exactly for the people who don’t want to mess with the technology but still do a presentable job. I’ve been able to produce some very nice recordings with mine, but I’m not a first-time user, and they won’t do everything.
Koz
Yes it does - use Ctrl+M to create a label at te current recording or playcack position.
WC