Uneven MP3 encoding quality

[Cutter]

Hello!

I noticed that the enconding quality of OC Remixes is uncertain. Some remixes are encoded in VBR, some other just stick to CBR @ 160kbps. Some sound pure, some other sound crap despite their musical quality (Time Chill, or Bonus Retreat for example). I imagine that not all remixers are aware of what is the best way to encode music into MP3. Shouldn't all remixes be encoded in a standard method? The remixer would send its uncompressed work (in zipped .wav format or in another lossless format), then djpretzel (or someone else) would encode it before releasing it on the site. Thus there will be no chance of disappointment finding a remix that rocks on the creative side but sounds crap because of compression artifacts.

Anyways, remixers should always at least encode their work with LAME which is known to be the best MP3 encoder to date.

You can download the actual recommended LAME version from here.

This encoder is to be used with recommended encoder settings.

Thanks in advance.

[sgx]

Sending in uncompressed files would just be bad. 1) They get craploads of submissions, so thats a lot of downloading for them to do 2) I'm sure some remixers are on dialup

A standard bitrate wouldn't be practical because there is already a 6 mb file limit. A three minute song should be allowed to have a higher bitrate, and a 6 min song should be at 128kbps in order to get the full length in.

[zylance]

Besides, my mixes usually are dramatic, as in the complexity changes much. I like VBR because it can really bump up the sound in the complicated parts while keeping the size low.

[Cerrax]

I use iTunes to convert my songs from AIFF to MP3. It goes up to 192Kbps and I can edit ID tags and everything.

[zircon]

This is impractical for a variety of reasons, not the least of which is that many submitters are on 56k and/or have no hosting space, which makes sending even a 3-6mb file difficult. Multiple that by 15? Uhh.. yeah.

[Cutter]

OK, forget about sending files in .wav format, but the remixers should be aware of the best method to encode their work. LAME should be recommended, VBR should be mandatory. It's a pity that there isn't a single topic explaining how to properly encode MP3s. Remixers may not be sensitive to compression artifacts, but a trained ear will suffer if the encoding is crappy (iTunes for example).

[Compyfox]

So what? Standard settings for LAME from 128kbit to 192kbit are totally fine. And most ears can't hear that out anyway (especially higher than 192kbit).

Also a track is only as good as it is produced. Encoding artifacts... can you really hear them especially on a portable device? Most of the songs from the last 2 years - either on OCR or VGMix - are perfectly fine. As long as it sounds good. And nowadays you can't do anything wrong anymore.

BTW:

AC3 (stereo) encoding is also 192kbit, non VBR (in 5.1 it's 448kbit). And AC3 is 48kHz 24bit - does that mean that it sounds bad too because it's CBR? We`re talking about more "information" and frequencies here.

[DarkeSword]

OK, forget about sending files in .wav format, but the remixers should be aware of the best method to encode their work. LAME should be recommended, VBR should be mandatory. It's a pity that there isn't a single topic explaining how to properly encode MP3s. Remixers may not be sensitive to compression artifacts, but a trained ear will suffer if the encoding is crappy (iTunes for example).

Trained ear? More like the ears of self-proclaimed 'audiophiles.'

I would think that it'd be the remixers themselves who actually have the 'trained ears.'

In any case, if you think it's a pity that there's no topic explaining how to properly encode MP3s, feel free to make one. You seem to know a bit about the subject, and it's better to contribute than to complain.

[Cutter]

So what? Standard settings for LAME from 128kbit to 192kbit are totally fine. And most ears can't hear that out anyway (especially higher than 192kbit).

Also a track is only as good as it is produced. Encoding artifacts... can you really hear them especially on a portable device? Most of the songs from the last 2 years - either on OCR or VGMix - are perfectly fine. As long as it sounds good. And nowadays you can't do anything wrong anymore.

BTW:

AC3 (stereo) encoding is also 192kbit, non VBR (in 5.1 it's 448kbit). And AC3 is 48kHz 24bit - does that mean that it sounds bad too because it's CBR? We`re talking about more "information" and frequencies here.

It really depends upon WHO listens to the MP3. You may find that it sounds fine, but someone who's used to the MP3 format will notice artifacts more easily. And everyone gets trained to notice artifacts sooner or later after listening to MP3s. It also depends upon your speakers / headphones. If you listen to remixes on your portable player outside (with ambient noise), you won't notice artifacts. But why penalize people with trained ears and/or good gear?

PS: yes, AC3 is bad. Worse than MP3 anyway according to this listening test.

[Compyfox]

God that discussion again... what was the page again from that... audiophile mp3 group?

You can't compare that and I tell you exactly why:

AC3 has a RMS limit, and that is -12dB plus 4dB for mezzoforte passages and explosions. That gives you a maximum of -8dB. Or better known as K-12 from the K-System invented by Bob Katz from Digital Domain. That's why it's perfect for DVD audio - you have access full dynamics.

MP3 however can go as high as -4dB. But it can't take anymore than -6dB without digital clipping - we proved that in another thread by SGX a couple of months ago. In terms of dynamics totally different.

Those tests were made by "selfproclaimed audiophiles", like Darke said - without proper metering tools. This topic came up at least 4-5 times and resulted in a huge flamewar where even the words fell "STFU you know jack shit about that topic" and the insults even continued on that board where this listening test is from.

Darke is right. It's up to the remixers to decide which format and encoder they use. It's a matter of preferences and opinions. It's not about gold cables (who barely "raise" the soundquality - only by 1%) or 2000bucks expensive dolby THX amps if you have a crappy room setup or can't even "hear" all frequencies anyway (which is only natural as human).

You can't force somebody to do it as "you" want it. Thathappened for the "Chrono Symphonic" project for example... You miss just one tag and the people freak out. You don't offer a FLAC version and the people freak out. What's next... another discussion that OGG is supperior than MP3 in terms of sound?

That's nitpicking and in my opinion this thread doesn't really lead to anything rather than insulting and preset minds/opinions on the issue.

Trust me... that won't take long anymore.

Anyway... I stick to LAME3.96 (VBR) for general material and own encodings, Nero AAC (mp4) for "online presentations" (if needed) and AC3 as standard for Dolby Digital productions (DVD). That's totally fine for me... and I'm an audio engineer.

[Cutter]

Trained ear? More like the ears of self-proclaimed 'audiophiles.'

Not self-proclaimed. If I claim that I can recognize artifacts in one of your remixes, I will prove so by taking an ABX blind-test. I will have to compare two files of the same music: one compressed, the other not. If I can recognize wich one is the original many times succesively, there would be no doubt that compression has damaged the sound quality, and that I can hear it.

In any case, if you think it's a pity that there's no topic explaining how to properly encode MP3s, feel free to make one. You seem to know a bit about the subject, and it's better to contribute than to complain.

The harder would be to convince remixers to adopt a method wich they may not be familiar with and for wich they might not see (hear) the immediate benefits. My english skills are probably too low for that.

[zircon]

To be fair, 128kbps vs uncompressed is very clear. Even 192kbps vs uncompressed on a set of headphones is pretty noticeable. I'd say ABOVE 256kbps is roughly the upper limit.

[Cutter]

God that discussion again... what was the page again from that... audiophile mp3 group?

You can't compare that and I tell you exactly why:

AC3 has a RMS limit, and that is -12dB plus 4dB for mezzoforte passages and explosions. That gives you a maximum of -8dB. Or better known as K-12 from the K-System invented by Bob Katz from Digital Domain. That's why it's perfect for DVD audio - you have access full dynamics.

MP3 however can go as high as -4dB. But it can't take anymore than -6dB without digital clipping - we proved that in another thread by SGX a couple of months ago. In terms of dynamics totally different.

Lol!

Those tests were made by "selfproclaimed audiophiles", like Darke said - without proper metering tools.

What better metering tool than the human ear? These tests aren't subjective: people didn't knew what they were rating. I won't try to convince you that these tests are reliable, this is off-topic.

[Compyfox]

Lol!

This is the reason why I said that this thread leads to nothing. You laughed about facts and standards, confirmed and set, by the AES (Audio Engineering Society) not to mention extensive tests.

For me this discussion is over, sorry.

[Cutter]

To be fair, 128kbps vs uncompressed is very clear. Even 192kbps vs uncompressed on a set of headphones is pretty noticeable. I'd say ABOVE 256kbps is roughly the upper limit.

It's not only a matter of bitrate. Using the proper encoder with the proper settings can make you save a lot of quality. If you find that 128kbps vs uncompressed is very clear, you probably didn't encode it the best possible way. The latest listening test from HydrogenAudio has shown that LAME can produce near-transparent (un-noticeable) sounding MP3 at this bitrate, if used with the best settings. Sorry to sound so boring with these listening tests, but that's the only way I know to show concrete facts. I don't say "I prefer LAME", I say "people in general prefer LAME". That's what listening tests are for.

[analoq]

let me take a step back from the technical discussion and try saying something.

if i understand you correctly, your problem is that some remixes on this site aren't encoded too great. that is true, but those remixes usually fall into two categories:

1. Older ReMixes

ocr had lower standards for accepting remixes in the past, and that can include the quality of the encoding.

these days if the judges get something that is encoded bad they get the remixer to reencode it if possible.

2. Long ReMixes

the 6 MB limit is hard and firm, no remix escapes it.

so if it a particularly lengthy piece of music, it has to be encoded at lower bitrates.

your suggestion that remixers should use LAME is a good one but it's redundant as we already recommend it:

Bitrate should be as low as possible to achieve the desired sound quality, within reason. We recommend and use the free MP3 encoder LAME. If it's a really short song, feel free to bump up the bitrate a bit, but do not exceed 192Kbps . . . if it's a realllly long song, and won't fit under 6MB, please consider cutting a shorter, alternate version. It's a bandwidth thang ^^

the infeasibility of submitting remixes in lossless formats (wav/flac) has already been pointed out.. so what is left to discuss?

that VBR should be mandatory? i don't think that's necessary.

cheers.

[zircon]

To be fair, 128kbps vs uncompressed is very clear. Even 192kbps vs uncompressed on a set of headphones is pretty noticeable. I'd say ABOVE 256kbps is roughly the upper limit.

It's not only a matter of bitrate. Using the proper encoder with the proper settings can make you save a lot of quality. If you find that 128kbps vs uncompressed is very clear, you probably didn't encode it the best possible way. The latest listening test from HydrogenAudio has shown that LAME can produce near-transparent (un-noticeable) sounding MP3 at this bitrate, if used with the best settings. Sorry to sound so boring with these listening tests, but that's the only way I know to show concrete facts. I don't say "I prefer LAME", I say "people in general prefer LAME". That's what listening tests are for.

I use LAME for everything, no need to convince me.

128kbps is still a massive difference for my ears. I do a lot of testing when I'm encoding my own music, and I found (like I said) that 256kbps seems to be the minimum for me to not hear a difference.

edit; keep in mind I'm talking about CBR not VBR.

[Cutter]

Lol!

This is the reason why I said that this thread leads to nothing. You laughed about facts and standards, confirmed and set, by the AES (Audio Engineering Society) not to mention extensive tests.

For me this discussion is over, sorry.

Sorry, I didn't get that you were serious about this technical slang I don't unerstand anything about. I thought you were simply trying to show that all this mp3 encoding stuff is uselessly complicated. My aim wasn't to hurt you at all.

[Compyfox]

You also miss something important (duh I`m posting again):

The lower the volume (loudness), the better the VBR encoding, the better the "sound" of the mp3. The higher the volume, the more squashed the sound (dynamics), the less it is possible for the encoder to convert it in a "transparent way".

Therefore the encoding is also in direct relation to the music/production. And if you laugh about that again (like with the post about the K-System and AC3), then you should start getting some background information before you "advise" us to change something. "Listening tests" don't really solid-out what you're trying to do here.

But that's finally it from my side.

*I was right from the start that this doesn't lead to anything*

[Cutter]

let me take a step back from the technical discussion and try saying something.

if i understand you correctly, your problem is that some remixes on this site aren't encoded too great. that is true, but those remixes usually fall into two categories:

1. Older ReMixes

ocr had lower standards for accepting remixes in the past, and that can include the quality of the encoding.

these days if the judges get something that is encoded bad they get the remixer to reencode it if possible.

2. Long ReMixes

the 6 MB limit is hard and firm, no remix escapes it.

so if it a particularly lengthy piece of music, it has to be encoded at lower bitrates.

your suggestion that remixers should use LAME is a good one but it's redundant as we already recommend it:

Bitrate should be as low as possible to achieve the desired sound quality, within reason. We recommend and use the free MP3 encoder LAME. If it's a really short song, feel free to bump up the bitrate a bit, but do not exceed 192Kbps . . . if it's a realllly long song, and won't fit under 6MB, please consider cutting a shorter, alternate version. It's a bandwidth thang ^^

the infeasibility of submitting remixes in lossless formats (wav/flac) has already been pointed out.. so what is left to discuss?

that VBR should be mandatory? i don't think that's necessary.

cheers.

Size being a big problem and VBR not being madatory is almost paradoxical IMO. The point of VBR is precisely to reduce filesize while keeping the perceived quality at the same level. Why using CBR @ 160kbps when you could get the same quality with VBR @ 128kbps? In other words, CBR is a waste of space and bandwith. Remixers should know how to control the filesize without having to reduce the quality.

[Cutter]

To be fair, 128kbps vs uncompressed is very clear. Even 192kbps vs uncompressed on a set of headphones is pretty noticeable. I'd say ABOVE 256kbps is roughly the upper limit.

It's not only a matter of bitrate. Using the proper encoder with the proper settings can make you save a lot of quality. If you find that 128kbps vs uncompressed is very clear, you probably didn't encode it the best possible way. The latest listening test from HydrogenAudio has shown that LAME can produce near-transparent (un-noticeable) sounding MP3 at this bitrate, if used with the best settings. Sorry to sound so boring with these listening tests, but that's the only way I know to show concrete facts. I don't say "I prefer LAME", I say "people in general prefer LAME". That's what listening tests are for.

I use LAME for everything, no need to convince me.

128kbps is still a massive difference for my ears. I do a lot of testing when I'm encoding my own music, and I found (like I said) that 256kbps seems to be the minimum for me to not hear a difference.

edit; keep in mind I'm talking about CBR not VBR.

You wouldn't have to use such a high bitrate as 256kbps to get transparent sounding music if you were using VBR. Like I said, it's a waste of diskspace.

Oh, and let me add that simple VBR doesn't produce the best results. The recommended settings for LAME that I gave in the first post gives even better quality (they are "optmized" VBR).

[zircon]

I DO use VBR. Check out "Clash at the Mountains" for example.

[Cutter]

I DO use VBR. Check out "Clash at the Mountains" for example.

Indeed. Did you use some particular command line like --alt-presets?

[Chavous]

Ok, I, personally, don't like overcompressing my music either, but let me ask you something.

Did you ever remix a song for more than two months?

Do you realize that by the time it is released/submitted, some people are so sick of hearing it over and over and over again that you just encode it so it is under 6 megs at a good quality and don't care because half the people on the site don't care either?

You get so sick of your own song, you just make sure it sounds good and nothing more. I've gotten sick of working on songs before, and I hate to hear it more than I need to. You want a set bitrate? You start a site that accpets only the top remixes at a sampled rate. The rest of the world doesn't really care, as long as it's not a horribly bad compression.

[zircon]

Not exactly. I use the Razorlame front end if I actually care enough, or DBpoweramp otherwise.