I’m an opera singer and have for a while been doing my own demo recordings with a Sony Linear PCM Recorder PCM-M10. I have achieved quite good results results doing the recordings in the “manual” mode and with low sensitivity in order to avoid clipping. Afterwards, I master the recordings in Audacity and adjust the volume.
Now, unfortunately, my Sony recorder broke down during an important recording. A friend agreed to come by with her equipment, so we were able to finish the session. However, listening to the files, I realize that she must have been recording in “automatic” mode: On the recording you hear how the microphone turns itself down in the loud passages - not nice. It basically sounds as if the recording has been “fixed” - and I, in contrast, need it to sound natural and unhampered with.
Does anyone know of a way to fix this?
I have tried to manually boost the individual passages but the result is not very organic.
Do you think a professional would be able to fix the problem?
Theoretically it is possible to fix with careful application of the Envelope Tool (http://manual.audacityteam.org/o/man/envelope_tool.html), but in practice it would be extremely time consuming and would probably still sound wrong. I tried this once and spent hours on it but eventually decided that the only solution was to re-record.
The conspicuous defect I can hear on that dropbox file is clicking (due to skipping) which is fixable using “repair” on each click …
The “automatic mode” problem [AGC] you describe , (which I can only just hear occasionally on “Sieglinde 30.11.WAV” ) , may be correctable automatically, to some extent, using an “expander” which does the opposite of AGC so can be an antidote. Chris’s compressor will act as an expander if you select a negative value for the compression ratio. Other expanders are available , e.g. “floorfish”.
I’ve never had success correcting AGC with an expander. Whenever I’ve tried to do that it has proved impossible to get the expander to track in synch with the previous compression curve, so I would not hold out much hope for automatic correction.
OK what about this method for automatic correction : use the level of an artifact , (e.g. a mains-hum harmonic), as a reference and expanding so it is constant throughout the recording, ( the assumption being it was constant amplitude and the variations are due to the AGC )?
We’re going a bit off topic from the original enquiry (which is asking for a practical solution before next week), but I guess there is some possibility there, provided that there is no actual audio at that frequency.
Trebor,
A most interesting idea.
I guess you propose that we should look at the noise floor to determine the actual used compression.
A normal recording should have a constant noise floor (hum, white noise, hiss).
This floor is naturally raised, if no gating was involved in the first place.
One could take an FFT and add up a certain percentage of the lowest values to estimate the noise floor.
For instance, the lowest 15 % (a common value for the noise floor) of the entire spectrum would be added, regardless of their frequency. You could now do an inversed automatic gain control, based on the individual deviation of all frames against the average noise level.
By the way, there’s a Nyquist plug-in available that does a compensation of AGC. I doubt that it will work in this particular case, since the compression curves will probably not meet.
thank you so much for all these inputs - I am going to carefully work through them together with my husband, who is an electrical engineer, in order to see if we can try out some of your ideas.
Trebor, you are right, clicking is a problem, too, but in the present situation this is less of a problem to me. I need the recordings to sound unedited - and for a sound technique amateur like myself, the microphone issue gives exactly the wrong impression in that respect: it sounds like I have been manipulating with the recording.
I used the Envelope tool. It took me about 10 minutes for just that one correction. So yes it can be done, but if there is more than a few minutes of recording then you will be having sleepless nights for the next week. If making a new recording is at all possible I’d recommend doing a retake.
A note that this isn’t a “help desk.” It’s a forum and you should post back if your friend figures out a way to do it – other than the way we did it
Koz
