I am a newish user of Audacity and am asking for tips in transferring informal recordings done on personal cassette recorders in about 1980, of talks given in a room to about 30 people. The microphone is not always that near the speaker and his level can be low eg -18 but then there are sudden peaks of his voice, or laughter from the listeners which overshoot to 0 and red.
I want to obtain the best quality of sound (low hiss, no distortion)at the best level I can given the equipment I have available (which has been lent for this task) - Ezcap Super USB Cassette Capture, onto an Acer laptop with Windows Vista. The settings I am using are Host: MME; record devices Microphone Array and Mono (and 2usb speakers for playback).
I am aiming for a peak of -6 but then a lot of the speech is at really low level.
I am not sure whether: to have the Ezcap vol at nearly max and reduce record level;
OR have the Ezcap vol less and increase record level.
AND I have tried out the limiter and also the leveller effects tools but these have to be done after the recording is that correct? In which case I don’t know if the limiter eliminates the break up noise of sound that has overshot, or if the leveller then affects quality eg. adds other fuzzy sounds underneath.
Any suggestions will be gratefully received.
To avoid distortion / clipping, the recording level must be set low enough that the loudest parts of the audio do not touch the top or bottom of the Audacity track. You can also monitor this by looking at the recording meter and ensure that it is always below 0 dB. (http://manual.audacityteam.org/o/man/meter_toolbar.html#recording)
Yes that is correct.
The Leveller adds distortion. My recommendation is Do Not Use the Leveler Effect.
If you record the audio too loud (the loudest parts hit the top or bottom of the track), then the Limiter will Not help to remove the distortion. However, if you have recorded low enough so that the loudest parts remain below 0 dB, then the Limiter may help to bring up the level of the quiet part without causing additional distortion to the loud parts.
I’ll move this topic to the Windows forum as you are using Windows Vista.
Please confirm that you are using the current Audacity 2.1.1 version (look in “Help > About Audacity”). If you are using an old version, please update to the current version: http://web.audacityteam.org/download/windows
There is only one critical setting. Once you overload the sound channel, the harsh, crunchy digital distortion and cracking is obvious and permanent. You should adjust the system so the loudest desirable sound in the show doesn’t go all the way to sound meter 0 (on the right) or 100% on the blue waves. Past that, record the loudest you can.
The painful decision is what to do about the laughter. How valuable is the show? If you’re being paid a million dollars, transfer the show twice, once for the lecture and once to get the best loudness on the laughter. Process the laughter as a separate show and lay it into the main show in place of the distortion. How many weeks do you have?
You can do that in place, too. Record for the presentation and then drag-select all the laughter overloads and apply filters to make the sound less harsh. We’re talking brute force like simple muffling. Turning the treble down, but only during laughter.
Or you can leave the overloaded laughter, and that’s just the way it is.
You’re going to have a very serious struggle with low-level noisy lecture and overloaded laughter. Normally, we would say you were too close to permanent damage and you have no show.
OK, that was the laughter. Now post a track without changing any settings with the voice. I would like to directly compare them as if they had come from the same show without stopping or changing any settings.
There is already some distortion during the loudest audience noise. Not much can be done about that - I’d just ignore it
If you want to use the Noise Reduction effect (a good idea in this case), it’s best to use it before you start messing about with the levels.
I’ve used quite aggressive settings of 18, 3, 3 (probably a bit too aggressive, but good enough for demonstration purposes ). For a more natural sound (but more hiss), use less aggressive settings (perhaps 12, 6, 6).
The “Soft Limit” setting in the “Limiter” effect can help to control the level. The maximum level during the talking is about -11 dB, so you would want to set the “threshold” a little above that- say about -10 dB. This will prevent the audio from going above the set level (-10 dB). I’d recommend setting the attack and release to maximum so that the level does not jump around too rapidly.
Although the Limiter does help, personally I don’t think it does enough. Ideally I’d like the loud sounds to be reduced even more than with the Limiter. There’s a plug-in that does something similar to this called “Pop Mute” (http://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Pop_Mute), but as the name suggests, it is designed for muting short sounds. However, as it is a “Nyquist” effect (written in plain text using the Nyquist scripting language http://wiki.audacityteam.org/wiki/Nyquist_Plug-ins_Reference) it is quite easy to modify the effect to suit our purposes. muteloudsounds.ny (1.19 KB)
This is a modified version of the “Pop Mute” effect. The main difference is that it allows the “attack time” (how rapidly the effect kicks in) and “release time” (how rapidly the effect returns to normal) to be much slower than in the Pop Mute effect. For audience noise, Attack and Release of about 2 seconds should be about right. The effect reduces the level of sounds that go above the “Threshold” level. Again we would want to set this a bit above the level of the speakers voice, but because it is a more aggressive effect it is advisable to allow a bit of headroom. I set the Threshold to -9 dB and reduced sounds above the threshold by -18 dB.
Thank you. I am trying things out. However I can’t open either of muteloudsounds.ny or processed-sample.flac. that you have posted. Have tried the latter in Realplayer and Windows Media Player and neither works.
I have 2 other questions:
About setting levels for the initial transfer with Ezcap: I don’t know which is best, to set vol control on Ezcap at nearly Max and record vol of about 0.52 to 0.60 or reduce the vol control on Ezcap and have a higher record vol. Would it make a difference?
About Normalize: when I tried it out at -1 having used noise reduction and limiter, everything was peaking at 0, in the red.
“processed-sample.flac” is in FLAC format. This format provides the same sound quality as WAV but in half the file size. It is an “open” format that has been well established for many years, but Microsoft have only just (in Windows 10) decided to support it. Virtually all good media players support FLAC format, including VLC, Foobar2000 and many others (I’m not a fan of WMP or RealPlayer). More importantly, FLAC is supported in Audacity, so just import the file into Audacity to listen to it.
It could make a difference. I’d suggest that you try reducing the level in Ezcap and set the recording level to max. That ‘may’ reduce the distortion on the very loud parts, but if it sounds worse, go back to your current settings.
Thank you. I’ve now been able to import and listen to your processed sample and put muteloudsounds into plug ins.
With record levels I tried Ezcap vol down and Rec vol max but I think this adds hum. Could this happen? I’ve tried a compromise with levels on both.
2 more questions:
Using Noise Reduction I seem to get peculiar effects even on low settings, and the same sort of thing happens with your sample. Possibly “out of phase”? The voice sounds odd, and there are added noises difficult to describe but like pops and what you get sometimes on a mobile.
Normalise or Amplify? I wonder what is best with mono and one track.
Yes, if you go too far one way you will get distortion, and too far the other way you will get noise (hiss / buzz). Compromise is the way to go.
Noise reduction is imperfect. There is always some degrading of the sound that you want to keep, so again it is a balancing act between maximizing removal of noise and minimizing damage to the remaining audio. The type of “damage” caused by this type of noise reduction is often described as “metallic bubbling”. Some sounds are more prone to this unwanted artefact than others… The more noise there is, the harder it is to remove. Ironically, noise reduction works best when the noise level is so low that it is hardly needed at all.