MPC is lossy compression, and because of this, there are instances were it fails. But, given ideal listening conditions and training, can a person be reasonably sure that they will never encounter a situation were audio compressed with MPC will be distinguishable from the original. And, if so, at which quality setting?

I'm not looking for an absolute guarantee, just a damn good one. For example, I know that people do die in airplane crashes, but I would be willing to trust the safety of my entire family to an airplane on a regular basis.

My objective is to compress audio at the lowest bit rate possible, but still be reasonably sure that no one will ever be able to hear the difference. So, not an absolute guarantee, but one good enough to stake my life on.

From what I've read, I'm guessing that MPC at Q6 would fit that description, but I'd like to get more input.

Thanks.

- Ian

There are rare cases at which q6 fails to be transparent (see this topic for details).

~ Florian

I think q7 (fails more gracefully than q5) is what you would be looking for or maybe wavpack hybird.

I guess it depends on just how worthless one's life has become.

There are no known lossy codecs with a 100% perfect reproduction of sound, guaranteed. The question is not could they fail, but when, and how bad?

With life on the line, the only sure stance is lossless (in a perfect, error free world...).

What you want is Q5 since it will give you transparancy at the lowest possible bitrate.

Refer to this faq:

http://www.personal.uni-jena.de/~pfk/MPP/a...nglish.html#faq



Will mp+/mpc reach much higher audio quality with forthcoming versions or is the encoder near the final state in terms of quality?

In terms of quality the encoder has almost reached the final state. The modifications with version-changes are mostly related to debugging of file-i/o and parsing and not to the encoder-kernel.
The current encoder has proven its reliability in terms of quality in several tests by many users and on some hundreds tracks. Only direct A/B-comparison using high-quality headphones and under hard effort of listening may show minor differences to the original with a small number of tracks. None of the encoded files showed heavy or annoying artifacts.

Will the "-standard"-profile do for encoding or should I use "-xtreme" or even "-insane"?

The encoder was tested intensively and optimized in "-standard"-profile, the default setting. In this mode the quality of the encoded tracks reaches - despite the profile's naming - very high level!
The next profile "-xtreme" uses slightly modified parameters to lower the quantization noise further below the masking threshold - it offers even more headroom.
For the "-insane"-profile the parameters are tweaked heavily. Using this mode will store the full bandwidth of the input signal and lead to much higher bitrates than "-standard" or "-xtreme" need. The storage of full bandwidth is not based on psychoacoustic reasons - it was implemented at some users desire.
Summarization: When using "-standard"-profile you will get high quality audio-files. If you want to push it a bit further use "-xtreme". The use of "-insane" is not necessary in general.

Pick a random track from a random CD
Encode using MusePack -q5, decode, burn original and decoded to a new CD.
Pick a random stereo system
Pick a random person off the street
Explain ABX to them
Ask them to ABX the two tracks
If they score "less than 5% chance of guessing", kill yourself.


I wonder how the chances of being killed in this scenario compare with the chances of being killed when crossing the road?


I bet you're more likely to die crossing the road.

But I have to cross the road. It would be silly to stake my life on MPC, because I know it isn't 100% perfect, and there's no reason to put myself in a position that its imperfections would cause my death!

However, I'm happy to listen to it.

Cheers,
David.

We'll miss you, 2BDecided...

You will statistically die in 5% of cases, as you yourself wrote in your setup

Shadowking, first, thank you for joining and helping,



Actually, yes, I found one encoded file with a heavy annoying artifact (to my ears) : http://www.hydrogenaudio.org/forums/index....showtopic=15505

But this is about 1 out of 1300.

LOL!! I nearly missed that, Pio!

I can't find a smilie with a big enough laugh!!!!

It makes you think though - that's 1 in 20. I can quite well imagine that for every "mpc just ABXed with this very difficult sample" thread, there are another 19 times where someone tried to ABX something and gave up because they couldn't. So maybe everything is transparent, and these "problem" samples are just the 5% ABX error rate?

No - I don't believe that either - but there are lies, damn lies, and statistics!


Surely it also means that, for example, 1 in 20 people should be able to pass The MAD challenge, even if the files were bit-perfect identical!?

That is food for thought. It could also impact the suggested "where is mpc transparent" listening test - is it deemed not transparent because 1 person can ABX - or is it only not transparent because more than one person out of 20 can ABX?

Waiting for that statistics genius again... I've just demonstrated that I don't quite qualify.

Cheers,
David.

I've taken to using quality 5 and am quite happy with it. The only thing I worry about is if i, at some point, want/need to transcode to mp3 for portable/car use. I know there was a transcoding test proposed/undertaken? Anybody have any clear cut results?

The way I see it, If i transcode for portable/car, these aren't ideal listening conditions and the evilness of transcoding probably won't be that apparent.

To get less than 5% you only need to make 5 correct guesses, remember! Thus the 1/20 chance that Pio brought up (its actually a 1/31 chance). If you said they had to try it 20 times and get less than a . . . Oh wait, I'm an idiot. 5% probability of guessing . . . . wow Definately 1/20 times.

MPC is a lossy compression scheme. The only thing you can be sure about is that it fails on samples where it is already known to fail.

Without a crystal ball, you will never be sure that a lossy encoder won't fail one day on a sample.

That's what this community is for, when someone finds a result, other can try and confirm it

...and what about my 50/50 with the FSOL sample ??

Well, you tried 1520 different tracks...

...but 50/50 should succeed only once out of 1,125,899,906,842,624 tracks

I won't be trusting my life to any audio compression codec. There are too many other things I have to trust my life to.

Air travel.
The brakes in my car.
Scuba gear.

Sometimes we need to step back and remember the point of all this stuff is to be able to enjoy music on a PC, portable or CD based track player of some kind. After all, even MP3 is transparent when done with Lame APS and storage just keeps getting cheaper.

Its hard for me to understand all the fuss over MPC when it is not under development, has no hardware support and the entire thing is in the hands of a quirky genius who would rather be doing something else in his spare time.

How come there's always someone from towards the AAC camp, trying to spoil the party

It seems we can generate quite enough motivation for this quirky genius again, if he has a new platform to code on. Not as bribe, but as an acknowledgement for his years of spare time work on MPC.

Would I trust my life to MPC?

Nope... I won't trust my life to any single thing on this world.

But then, nothing in this world is perfect and trustable. We just take whatever is better for our need. Instead of asking whether you can trust on a lossy format or not, probably you should ask how much can you trust on the whole music production and reproduction process? Can you trust on the microphone used to record the music? Can you trust the audio engineer who mixed the recordings? Can you trust the media that delivered the music on? Can you trust the equipments that the music will be used to play on? Can you trust the noise surrounding your environment that will not mask your hearing ability? Can you trust your ear and brain that will interpret the music faithfully? .....

MPC has been rigorously tested and found to have less problematic samples than other lossy formats. Right now it is the best lossy encoder. It produces near transparent music for lowest bitrate possible. But still there are few (very few indeed) samples that it will fail to encode properly. If you are planning to use Q6 then 99+% music will be transparent. Some people tend to use Q7 just to give a breathing room for transcoding purposes. In few cases it gave better result when transcoded to another lossy format. But for most people, Q5 or standard setting is enough.

That is highly arguable. I never saw any proof of that. Actually, ScorLibran is planning a test exactly to check that.

Actually there are quite a number of reasons why there are that much fuss over MPC. It is transparent at the lowest bit possible, has the best psychoacoustic model implemented, it can play gaplessly, has one of the best tagging format support, replaygain supported and most importantly it is tested rigorously in public and most problems have been taken care of. No other single lossy format can claim to have all these important features at the same time.

The issue that there is not much development recently is less of an issue for most of the users. Because it is already best available format on the market. Even if it is not developed anymore it won't be less attractive until other format can surpass it in terms of features. Moreover, nobody has said there will be no further MPC developments.

The point is no single format will be developed for life. People will adopt whatever format serve their purpose best for that time being. Granted it is not for everybody. It is only for those who seeks the featuresets it offers.

Again, you are claiming something without proofs. Rule 8 anyone?

Enough tests were done before to believe that. Yes, those tests are outdated now. But no test showed this claim to be wrong. In a similar fashion, we claim that that LAME -aps is the best settings to use. The issue is not that any other settings (tweaking) could not surpass LAME -aps settings. It is just that it has been tested rigorously and more trustworthy.

Scorlibran's test will definitely help to reassess this statement.

Where?

Use search button...

We are again going off-topic here...

The burden of proof lies on you, not me.

Heh, that question depends on how you define "near transparency". For most people HE-AAC with 64 kbit might be the answer...

In my tests, after encoding albums in Lame APS and MPC -q5, as well and APE and -q6, I find that in all cases MPC beats the snot out of MP3 as far as file size. Not saying it's the smallest possible bitrate, but it's the most efficient for transparent encoding that I've encountered.

@citay

The only camp I am from is the reality camp.

Some benefits of MPC no longer being developed:

1. It is still the best lossy encoder (flame away).
2. It can't get any worse.
3. You don't have to re-encode your whole collection with each new version.
4. You don't spend a year (or more) doing quality tests with each new version.
5. You can just use it and not have to keep up with changes in development.
6. Last tweaked by a "quirky genius".
7. It is still the best lossy encoder (for now).

Why feed trolls?

New versions produce identical output, or what!? What's new then?


Not being actively developed.


Same argument.



I agree with Roberto's previous post here

Got one !!

@rpop

You do realize that it was all a joke, right?

Reel him on in!