Free Plugin or Filter for Audio Hijack to Reduce Artifacts From Low-Quality Audio?

January 20, 2005 8:44 PM

Is there a free VST- or AudioUnit-compatibile plugin designed to remove or reduce compression artifacts from low-quality audio being re-recorded at a higher bitrate? Or special settings for a specific filter that will do this? I need one for AudioHijack on OS X.
posted by Mo Nickels to computers & internet (6 comments total)

I don't know if I have an answer for you, but could you clarify; by "lower to higher" bitrate, are you meaning bitrate in terms of, say, a "22khz --> 44khz" sampling rate increase, or bitrate in terms of a "64kbps mp3 --> 192kpbs mp3" conversion?
posted by Jimbob at 10:19 PM on January 20, 2005

(Silly me - I meant to say "8-bit to 16-bit" sampling resolution conversion instead of the 22khz sample rate conversion).
posted by Jimbob at 10:19 PM on January 20, 2005

Once it has been overcompressed the information is gone. That said, what are you trying to do?

Adding white noise (the process is called dithering, just like with an image) will cover up compression artifacts very effectively, but you will hear the noise.

It is just possible that you are referring to dynamic range compression (the process of making loud passages and quite passages in the original come out at the same volume after the compression). Dither wouldn't help in this circumstance.
posted by Chuckles at 10:58 PM on January 20, 2005

yah, not much you can do with something like that.

If there is a specific sound that is most intrusive, you can just try to EQ it out with a notch filter.
posted by Espoo2 at 11:29 PM on January 20, 2005

What it is is a simply low quality 22KB RealAudio mono stream that I record to the drive as a a 64KB stereo stream. There are lots of compression artifacts at all frequencies. They sound vaguely like echos but are incomprehensible and tend to have a bit of a micro-flange effect.

Notch filters don't work, nor does EQ. I've tried all different sorts of high and low band pass filters, too. The artifacts appear across the frequency range. Dithering is not useful. I am not referring to dynamic range compression.

I know the information is gone, since that's the basic problem with most audio compression, but it occurred to me that if the artifacts were added by compression--which uses a known formula which means that given a specific input, it should always yield a specific result--that the compression artifacts would be predictable. Therefore, if they are predictable, then it seems like they can be recognized as such after compression--even without comparision to the original uncompressed stream--and removed. This particularly seems likely since there's always a bit of delay on the artifacts, and because certain frequencies--such as those used by high-pitched voices--yield more of them.
posted by Mo Nickels at 8:55 AM on January 21, 2005

The compression/decompression algorithms are already doing everything reasonable to account for the lossyness of the recording. It is said that there are "better" encoders/decoders for certain lossy compression algorithms (I've noticed this with mp3). Somebody else might know better, but my guess is that because RA is proprietary there is no chance of a different decode filter.

Dither will help your 22k RA stream sound better. The noises you are hearing are correlated to the audio signal, and that makes them very noticeable and irritating. Adding the dither will break up the correlation. Don't think of dither as just masking the problem.

On the other hand, dither doesn't compress well at all, so once you recompress at 64k you might actually be doing more harm than good (you probably are doing more harm, actually).

Compressed audio is bad! Monkey's Audio, flac, and shn are good!
posted by Chuckles at 1:41 PM on January 21, 2005