I have an on-going audio digitization project. The material has noise reduction issues. I’m looking for ways to simplify the handwork.
This material is old, but valuable. Most of the recordings have a lot of hum and noise - nowhere near enough to make it unlistenable, but annoying. It’s easy to make listening much more comfortable with very light noise reduction.
Lectures & musical tracks are not much of a problem. Often noise reduction can be skipped, it’s not obvious enough to most people to be a problem. I suppose it might be subliminally an energy drain on the listener, but don’t know.
Guided meditations make up a lot of the content, and they are a problem. The normal pace of these is often
Talk,talk,talk, <silence for 30-300s>, talk,talk,talk, <silence for 30-300s> &c
During those silent periods, the fluorescent induced hum, the crackling, and buzzing just jump out because of how the material is used - this noise really needs to go away. I’ve found that I can either de-amply the sound in the gaps by -50db or do noise reduction at a very high level, effectively silencing the non-talking period; either work well.
The problem is the hand labor with these guided talks. Is there a way I could select more than one segment in a track & apply this reduction to all of them at once? A way to script this? A way to apply high level NR to all sections of the track that are below some db limit? Some other idea? Any ideas appreciated.
Which Audacity are you using? You may be working too hard. Analyze the noise. If it has a strong power line component, you might be able to apply a hum remover followed by medium noise reduction to the whole show.
You might be a candidate for a Noise Gate. 'Watch the sound level and reduce the sound even further if the raw show drops past a certain loudness. Many people think you need to crush the quiet parts to zero, but these tools can be used more gracefully at smaller values. Hum Remover, gentle noise reduction and a small value of noise gate may do it for you—without programming or scripting.
Macros can’t make decisions. They can’t go looking for errors or gaps in the show. There is no IF/THEN. IF the voice drops out, THEN do processing. Sorry. It can’t do that.
Post some of it. If it’s mono, one blue wave, you can post up to 20 seconds of perfect quality WAV. If it’s stereo, you’re limited to 10 seconds of WAV. Stay away from MP3. MP3 creates sound distortions and you can’t stop it.
Include two or three seconds of noisy gap and the rest dialog or show.
Scroll down from a forum window > Attachments > Add Files.
Audacity 3.02 on Windows 10
I have encountered the “uncanny valley” feeling of having all the sound gone, but it’s better than the buzz.
Even the ability to select multiple segments & then apply the same NR to them would be a big work flow improvement.
As far as coding goes, I was wondering if something like
if ( db < some value over some time) ) apply deep NR
I attach a snippet as requested from one I worked on today. It’s far from the noisiest but it’s over the edge of where I would suppress the hum, & the hum is typical.
I’ve gotten hypersensitive to the hum in the lectures &al but I don’t think most people really hear it. On the other hand the gaps / quiet periods, it’s really unwelcome.
I think you win the prize. The clip has tape hiss (no surprise there) and European power buzz (50Hz and harmonics) and also “Stuff.” It’s the Stuff that’s going to be a problem. I can’t identify it enough to know the best way to suppress it other than, as as Jademan, above, the blunt force tool of Noise Reduction.
So I rolled up my sleeve and generated a recipe of tools. No one tool is taken to extremes, so the voice survives unharmed. You can still hear some hum between words if you search for it.
This has the advantage that you run the tools one time at the beginning no matter how long the show is.
Effect > Amplify is pretty handy. If you run it on some work, the bottom window defaults to 0dB (maximum volume) and the top window will tell you the boost needed to get it there. If you write down that number and add “-”, that’s the peak value of the work unless you change it.
The peak value of that sound is -16.021dB.
I think there’s a limit. It will only boost or amplify 50dB before it gets tired.
But you can apply it more than once if you need to find the value of something really quiet. Add the numbers from each pass.
I have just found one horrible track where it made a bad situation worse; what I (so far) found most effective was to skip most of the tools, fool around with noise gate, & just suppress everything in the silent gaps like before. Maybe I’ll post a small sample if you’re interested, from the original. The whole sequence made the original poor quality audio unintelligible in spots. Which tool do you think has the most effect on voice quality? I didn’t have time yet to try each one separately.
What I think is the most like what I was looking for is the noise gate. I’d like to understand this more. This seems to get rid of a LOT of unwanted noises including background in the talk &c without damaging the recording. I’d have to look for some musical tracks to see what happens there. But for tracks that aren’t too bad it cleans up the whole thing w/o having to dip & dive. The other tools are useful but this one effects improvement I didn’t have any approach to.
I have noticed that a few tracks still seem to benefit (to me) from more audio boost than this sequence provides. I am realizing I don’t understand the amplify tool very well either.
Taking this backwards. You are suffering from “The Software Should Know…”
The software doesn’t know anything. All Amplify knows is that somewhere in your hour long presentation, one tiny bit of blue wave was higher than all the others and it latches on that as the reference for the whole show. If that one little wave happened to be noise, then it will push the desired voice into the mud to accommodate that wav. It has no idea what “noise” is.
There are certainly ways to set loudness, the audiobook people have that, but it tends to fall apart if you’re in rescue or disaster mode.
The long, involved rescue I posted was intentionally designed so one single tool didn’t dominate. You can set Noise Reduction for much higher corrections, See: Jademan, above, but then the voice will start sounding hollow, wine glassy, and talking-into-a-milk-jug. You may decide that Noise Gate is your go-to tool and it might be, but high gate settings can cause upcut first word sounds and/or word noise tails and used alone, it can leave high volume background noise between the words.
You can certainly have a tape so damaged that the tools stop working. That may be the tape you correct manually and that may be the tape you decide you can’t fix manually, either.