plugin to fight mp3 encoding artifacts
-
- Tunesmith
- 2889 posts since 12 Mar, 2002 from Toronto
this recent thread had me thinking about an issue that comes up often - mixing your music so that it doesnt get beat up too hard by the mp3 encoders. has anyone thought of a plugin that you can put on your master chain to help overcome what the encoder will do to your music?
-
- KVRAF
- 1884 posts since 9 Feb, 2004 from Rochester, MN
If there was a simple plugin that could be added at the end of a mastering chain to reduce compression artifacts, shouldn't an MP3 encoder just automatically perform that bit of DSP while it's encoding?
Point being, if it's not in LAME already, then I doubt there's any universal process that would magically make MP3's sound better. On a song-by-song basis, you may be able to improve the quality if you mix specifically with an MP3 in mind, but I doubt it's really worth it.
Point being, if it's not in LAME already, then I doubt there's any universal process that would magically make MP3's sound better. On a song-by-song basis, you may be able to improve the quality if you mix specifically with an MP3 in mind, but I doubt it's really worth it.
Last edited by Warmonger on Sun Sep 24, 2006 7:29 am, edited 1 time in total.
- KVRAF
- 6504 posts since 25 May, 2002 from Bobo-dioulasso\BF__Geneva/CH
interesting subversive question, mate
As far as i can figure out, it should possible to concieve one or some algorithm that may minimise one or another of these artifacts, this of course with a relatively high quality compression ratio
By comparison with something existing...
( voxengo soniformer applied to compression, for instance )
...i can figure out that some specialised analysis/resynthesis algorithms should be the only approach
-BTW supposing you where opening this thread for another purpose than something like
i would say this would make sense applied to a track included as a kind of submix inside a mix,
I think it doesn't make sense to hope something that should work applied on a final mix because i suspect that theses issues where already taken in consideration by the devs who created the mp3 encoding systems...What do you think of this ?
As far as i can figure out, it should possible to concieve one or some algorithm that may minimise one or another of these artifacts, this of course with a relatively high quality compression ratio
By comparison with something existing...
( voxengo soniformer applied to compression, for instance )
...i can figure out that some specialised analysis/resynthesis algorithms should be the only approach
-BTW supposing you where opening this thread for another purpose than something like
I think it doesn't make sense to hope something that should work applied on a final mix because i suspect that theses issues where already taken in consideration by the devs who created the mp3 encoding systems...What do you think of this ?
-
- KVRian
- 520 posts since 11 Sep, 2003 from Berlin, Germany
MP3 itself *is* minimizing the artifacts. That's what perceptual coding is all about.
If you set the bitrate too low there's nothing you can do - except going mono or hi-cutting (which encoders do already).
If you set the bitrate too low there's nothing you can do - except going mono or hi-cutting (which encoders do already).
- KVRAF
- 10286 posts since 17 Sep, 2004 from Austin, TX
Pop it in your DAW once it's already an mp3 and monkey with it in there.
That should teach you something about what you're attempting to do.
That should teach you something about what you're attempting to do.
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
I work for a multimedia company, and when I know that the final output will be, say, Flash at 32kbps, then I will often mix in mono and have a lowpass filter at 3200hz in the main bus, and just flip it on and check it once in awhile. If I'm very concerned, I'll export a little 32kbps clip, and you can hear where the real problems are going to occur.
But chances are you want to play the mix at some point in an uncompressed or HQ format...so you might as well mix like normal, unless it's only ever going to be for the web.
But chances are you want to play the mix at some point in an uncompressed or HQ format...so you might as well mix like normal, unless it's only ever going to be for the web.
-
- KVRAF
- 3964 posts since 31 Aug, 2003 from In a foreign town, in a foreign land
Ogg Vorbis.
Groet, Erik
Groet, Erik
Pop music delenda est.


-
- Tunesmith
- Topic Starter
- 2889 posts since 12 Mar, 2002 from Toronto
Hmmm it seems like i'm getting a lot of answers that suggest "dont even go there"... but for starters i have been hearing about mixing engineers that have some tricks for mastering tracks that will be hitting mp3 encoders. I just dont know what their tricks are.
But I am referring here to tunes that will be imported into flash sites eventually so ogg vorbis is not an option.
Oh well i thought it was worth asking about at least
Thanks everyone for responding...
But I am referring here to tunes that will be imported into flash sites eventually so ogg vorbis is not an option.
Oh well i thought it was worth asking about at least
Thanks everyone for responding...
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
Weird...I'm trying to picture a $100/hour mixing or mastering engineer, huddled around their expensive equipment, "mastering" for...Flash. Ridiculous.Mr. Tunes wrote:Hmmm it seems like i'm getting a lot of answers that suggest "dont even go there"... but for starters i have been hearing about mixing engineers that have some tricks for mastering tracks that will be hitting mp3 encoders. I just dont know what their tricks are.
But I am referring here to tunes that will be imported into flash sites eventually so ogg vorbis is not an option.
Oh well i thought it was worth asking about at least
Thanks everyone for responding...
Mr. Tunes, in my humble opinion, I think you're hearing urban myths. I've been doing audio for Flash for 6 years, and I've never heard of such a thing. You did actually get lots of good advice about what the only thing you can do: test your music in the compressed format.
-
- Tunesmith
- Topic Starter
- 2889 posts since 12 Mar, 2002 from Toronto
Well you know what would be interesting - a plugin that emulates 128/kpbs mp3 encoding on your master bus then to do quick checks on how it would sound.bduffy wrote: You did actually get lots of good advice about what the only thing you can do: test your music in the compressed format.
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
Yeah...apart from the artefact flutter, and I assume that's your focus, you realize that bitrate is directly related to frequency; so just cut at 12.8khz and you'll get an idea.Mr. Tunes wrote:Well you know what would be interesting - a plugin that emulates 128/kpbs mp3 encoding on your master bus then to do quick checks on how it would sound.bduffy wrote: You did actually get lots of good advice about what the only thing you can do: test your music in the compressed format.
- KVRAF
- 6478 posts since 16 Dec, 2002
whooops. no. that's not how the MP3 format works at all. (it's rather a several magnitudes more involved)bduffy wrote:bitrate is directly related to frequency; so just cut at 12.8khz and you'll get an idea.
Great articles out there, but know that it's based on FFT and some pretty clever psychoacoustic masking algorithms at the core.
I'd stop worrying about encoder artifacts, and just use quality encoders, like latest LAME codecs at a high enough bitrates. For me 192kbps is the absolute bare minimum, but I can't remember encoding anything below 200kbps for years now.
I would love to use ogg for all lossy encoding, but it has a severe issue with compatibility. it's not that it's not supported, just the fact that vast majority of people have no idea how to get it or use it. The situation is worse on macs.
Last edited by Kingston on Sun Sep 24, 2006 6:05 pm, edited 1 time in total.
-
- KVRian
- 943 posts since 15 Mar, 2005
despite the very priciple of mp3 being its perceptual coding, there is actually one very simple thing you can do, leave headroom! the bit reduction and filtering creates peaks, getting bigger the lower you encode. if your tracks rms is hot and peaks solidly at 0dbfs, then you can expect new peaks of up to 3db at around 128kbps encoding, which will simply be digitally clipped and add to the distortion. v simple to test too.
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
I didn't meant that the mp3 codec works by lowpass filtering! I seemed to recall that the bitrate worked out to the frequency that'll pass through, I don't recall the math, but a conversation that I had with another multimedia guy few years ago. I may be guilty of hearsay also. There's instant kvr-ma for you.Kingston wrote:whooops. no. that's not how the MP3 format works at all. (it's rather a several magnitudes more involved)bduffy wrote:bitrate is directly related to frequency; so just cut at 12.8khz and you'll get an idea.
Great articles out there, but know that it's based on FFT and some pretty clever psychoacoustic masking algorithms at the core.
I'd stop worrying about encoder artifacts, and just use quality encoders, like latest LAME codecs at a high enough bitrates. For me 192kbps is the absolute bare minimum, but I can't remember encoding anything below 200kbps for years now.
I would love to use ogg for all lossy encoding, but it has a severe issue with compatibility. it's not that it's not supported, just the fact that vast majority of people have no idea how to get it or use it. The situation is worse on macs.

