Fixing audio recorded too low

Billy d wrote on 5/1/2007, 11:56 AM

Basically you are screwed. If you don't have the sound there, then there is no way to create it now.
Usually the most obvious degradation from an under recorded sound track will be an increased noise floor most noticable as hiss.
Try some eq to remove some of the extra hiss. It will be a compromise because you will lose the upper frequencies of the stuff you want as well.
Not sure why the bass is thin, low record volume shouldn't affect the bass.

If the two recordings butt up against one another (so that the difference is real noticable) then try a mix of:
1) Making the good clip get gradually worse up to the crossover and, on the bad clip, cutting the hiss out (by decreasing the gain on the high freq) and then gradually reducing the hiss removal after the listeners ears have adjusted, so that the higher frequencies of the stuff you want are not lost for the bulk of the recording.
2) Applying a "decoy" sound or visual effect at the crossover so the listeners attention is distracted from the drop in quality.
3) some combination of the above.
Good luck.

nolonemo wrote on 5/1/2007, 12:58 PM

Thanks for the transition ideas, Billy, that will make the transition less noticeable. Since there were two parts to the performance, I may fade to a title and back in to put some distance between the two clips.

There really isn't a noticable increase in hiss (the floor of the audio is really clean), and I can help the thin bottom with some EQ. I'll live with the crowd murmur, it's present in the first part, just not as noticable.

riredale wrote on 5/1/2007, 4:26 PM

Compress the first part in order to bring up the crowd murmur to match the second part.

JoeMess wrote on 5/1/2007, 5:22 PM

I would not just ride the gain control up. Normalize to a peak value so it matches the first half of the show. At that point, compare the quality. Keep the two halves as separate tracks, so you can experiment with plug ins on both halves separately and keep things organized in your mind as to what you did to each half. If you lost bottom, because you lost a direct feed and ended up with Mic only input. You will have some serious issues. I would strongly recommend using a multiband compressor and really punching up the bass frequencies and compressing the crap out of them so they are brought up and infront as they are in the direct feed. Direct Box bass is really hard to reproduce with an ambient room recording. A multiband compressor will let you get somewhere back in the ball park.

Joe

JoeMess wrote on 5/1/2007, 5:28 PM

OK, Sorry, I see this was a string ensemble. (Quartet, Octet?) This may be a job for Sound Forge. If you are fighting the noise of a classical crowd. (Program shuffling and coughs, etc.) You will probably want to draw out the noises in Sound Forge. I would still normalize to a peak value, but I would not use compression. (It will ruin dynamics and the performers will cringe when they hear it.) Scrubbing out the offending noises and normalizing to a commen peak value should get you buy.

Joe

JoeMess wrote on 5/1/2007, 5:28 PM

OK, Sorry, I see this was a string ensemble. (Quartet, Octet?) This may be a job for Sound Forge. If you are fighting the noise of a classical crowd. (Program shuffling and coughs, etc.) You will probably want to draw out the noises in Sound Forge. I would still normalize to a peak value, but I would not use compression. (It will ruin dynamics and the performers will cringe when they hear it.) Scrubbing out the offending noises and normalizing to a commen peak value should get you buy.

Joe

rchristiejr wrote on 5/2/2007, 10:05 AM

Do you have a Noise Reduction plug in?
You could seperate the signal from the noise.

What kind of noises do you hear?

nolonemo wrote on 5/2/2007, 10:30 AM

I don't think plain old noise reduction would work, the crowd noise is too variable - its a conversation of shuffling and low talking (this was not a concert situation, but a silent auction, so people were hanging out at the opposite end of the ballroom where the ensemble was playing, and some of the "audience" were chatting during the performance anyway. I had the mic close enough to the players that the background noise was pretty much drowned out by the music (mostly uptempo stuff) during the first part of the program.

TorS wrote on 5/2/2007, 11:57 AM

Billy d's suggestions are very good. But do make an audio transition, not just a visual one.
In addition: With EQ, lift the dominant frequncies ever so slightly. Then with Graphic Dynamics (or Wave Hammer if you have it) add a little compression to the track. We are talking subtle here. Add the same amount of compression (or a little less) to the good track too. If the EQ make a big difference, add it gradually towards the end of the good track, in keeping with Billy's thinking.

Do not be afraid of EQ or compression. But don't overdo them either.

(Audiences are always more noisy after the break, returning from the bar.)
Tor

dsf wrote on 5/4/2007, 10:34 PM

Would someone explain what is meant by "compression" of audio in this thread?

Also, why would neolemo not want to use plain old EQ to improve the bad sound on his second section? Or should I understand he DID use EQ but the sound was still not good enough?

Steve Mann wrote on 5/4/2007, 11:15 PM

I downloaded some software that starts the description with: "Do you believe in magic?"

My answer was no, so I parked the download in the folder where I keep things that I might try to use someday.

The software is "Levelator" from www.gigavox.com. And it's freeware.

So I shot a play that the sound feed from the mixer board was down 30dB from our test levels before the performance. No way to fix it now, the play had started. I tried boosting it in Vegas and it never sounded right. Well, nothing to lose, so I rendered my soundtrack to a WAV file and fed it to the Levelator.

I was expecting the noise to be raised with the sound, but as their website said - it really was magic. I was blown away with the quality of the output file.

Steve Mann

craftech wrote on 5/5/2007, 4:24 AM

Have you tried lowering the gain on the FIRST clip to match the second?

John

johnmeyer wrote on 5/5/2007, 8:26 AM

Would someone explain what is meant by "compression" of audio in this thread?

Short version: Makes soft sections loud and loud sections softer.

Longer version: Dynamic range compression - From Wikipedia

Wikipedia is your friend: If you don't know the answer to something, go there. If that doesn't work, try Google.

Steve Mann wrote on 5/5/2007, 9:30 PM

John, your description is of normalizing. Compression just works on the loudest parts of the audio.

johnmeyer wrote on 5/5/2007, 9:34 PM

John, your description is of normalizing. Compression just works on the loudest parts of the audio.

Thanks, Steve. I stand corrected.

douglas_clark wrote on 5/7/2007, 5:26 AM

Sorry but John is right. Compression makes the loud parts softer so that the overall level can be raised to make the soft parts louder. That compresses the dynamic range.

Normalizing raises the overall level of the track/event until the highest peaks are as high as they can be without going over 0 dB. That doesn't change the relationship between loud and soft parts as compression does. It just raises the volume as high as it can go without distorting the peaks.

Steve Mann wrote on 5/7/2007, 10:12 PM

That's what I said - compression just works on the loud parts. Then, as you said, you can raise overall level to recover the quiet parts.

Steve Mann

Bill Ravens wrote on 5/8/2007, 5:03 AM

there are 2 routes i would use, perhaps both.
1-In Sound Forge, only, use NORMALIZE with the RMS option set and Compress Peaks if 0 dB set. I would also explore selecting the noise gate that's a part of this normalize function. Note: don't look in Vegas, it doesn't do the same thing.

2-I've mentioned, before, a great program called HarBal. This is basically an EQ carver. With this tool, you can restructure the whole EQ curve for your recorded sound, eliminate frequencies with notch filters, and pretty much clean up any poorly recorded audio.