The Russian music site allofmp3.com say they offer their music encoded with lame presets as an option - standart (standart=standard in Russian), extreme, insane, as well as other formats.

The source for *most* music on allofmp3 is mp3 384kbps (some music has lossless as source for twice the cost), which I understand is close to lossless.

Is it possible to use the 384 kbps as source to create alt standard mp3?
Isn't this transcoding? If you start with MP3 does it make any sense to re-encode with lame?

Yes, it is possible. Yes, it is transcoding. Yes, I use it and it sounds very good.

It is possible to re-encode from free-format MP3s, it is transcoding, and it might sound good or bad - it depends on your ears.

As far as I understood, foobar (which is my player of choice) is not able to decode freeformat mp3, like the 384 kbps offered by allofmp3. So how do you aktually "transcode" that stuff? Decoding to wav with lame and then use lame again to create a standard mp3 format, or is there a more "direct" way to do it?

It's not you who transcodes the stuff, but ALLOFMP3.

By the way, LAME can create free-format MP3s, but cannot decode them back to WAV.

And this is because LAME and foobar use the same mp3 decoder AFAIK.

IIRC, MAD is the only decoder being capabile of decoding free-format MP3 bitstreams. LAME and foobar2000 both use mpglib.

I think it can.

Nope, it can't.

Well, mine can. (3.97)

OK, tested with 3.97 Alpha and it works, indeed. I was using the (still) recommended 3.90.3.
By the way, did you change the decoder to MAD in the meanwhile?

If I were gonna download from all ofmp3, i'd take the freeformat stuff. Then I could decide to do whatever I wanted with it.

No, and this is not planned. We would like to use an LGPL decoder, not a GPL one.

I am a bit confused now... How come it can decode freeformat MP3s, then? I thought MAD was the only decoder being capabile of doing so.

Then you were wrong

-Eugene

Not to our ears! We (a panel of self-confessed audio fanatics) listened to allofmp3's 384kB source files transcoded to LAME extreme standard (via WAV) and compared them with files (same material) ripped to WAV from our own CDs and then transcoded to extreme. So the only difference is the original source file. Guess what? We all agreed that allof mp3's files sounded dull and lifeless in comparison to self-ripped and encoded files. Co-incidentally, we found that the yawning quality gap between allofmp3 sourced files and our own material narrowed considerably when we listened to their LAME extreme files encoded from original PCM data (their exclusive service). We concluded that for anything other than portable listening on poorer mp3 rigs, allofmp3's standard files were not good enough!

So there

Was it a blind test?

...the problem about allofmp3.com is that they use very bad settings for ther --freeformat encodings! First of all, they use LAME 3.93.1 and secondly they use
--freeformat -k -b 384 -q 0 -m s ...someone try to explain to them that they should use --freeformat -b 384 ...I guess they don't care

Thank you for this information, minty. Can you name two or three of the songs where the quality gap was mentionably large (you mentioned Madonna's ray of light), secondly how many people where doing the listening test, and thirdly, as already requested, whether it was blind (I guess so). Thank you. Moreover, please tell me if it is the case that the quality problems may or probably are the result of the mentioned lame switches and versions they use. I simply wonder what it is that causes the quality problem...!

have you read my post

The test was blind and the track sources weren't revealed until the randomised listening tests were completed. It's quite possible that allofmp3's choice of LAME preset for freeformat might have something to do with the poor quality of their files, but I doubt that it would account for the obvious difference we heard.

Try this experiment; download a couple of standard freeformat source files from the Zero 7 album, 'When it Falls'. Decode to WAV and then recode using LAME standard. Now download the same files only this time use their exclusive service and repeat the decode/encode steps. Now listen to the mp3s. You don't have to listen particularly carefully to hear the difference.

The listening tests we carried out earlier were randomised and blinded.

Yes, I have. :-) As minty states, the (seemingly) obvious quality differences he and his colleagues observed are not likely to be caused by the somewhat strange freeformat settings allofm3 used (verifying whether or not the freeformat switches were the case would be easy btw). So in fact the question indeed remains: "what, then, was the cause." Of course you are welcome to point me to information proving these 384kbps freeformat settings yield mentionably bad results, worse than lame standard for instance.


:-)

However, this test could be flawed. How can you be sure it's the same audio data you are encoding from the source? The data of a music CD varies between countries, and there's also remastered and non-remastered versions of the same discs....

I will gladly do so, but first let me ask why it is that I need to reencode, I thought the problem lies in the source. Hence listening to the 384-version (or the MAD-decoded wav thereof) and then listening to the lossless "exclusive" version should be the quasi the same, shouldn't it? Transcoding from 384kbps files shouldn't be the problem. We're talking about quality problems that a half-way normal person ;-) should be able to distinquish. I'l do a listening test as soon as the above short question is settled.

Regards

m

why not download it directly as --aps? First you choose the 384kbps file as the source (1 cent / megabyte), second file you use their oeex (from a lossless file, 2 cents / megabyte).

Lame supports decoding of freeformat mp3's up to 550 kbps. MAD is the only decoder with full support for freeformat, i.e. 320 kbps - 640 kbps.

I suggested the decode/transcode just to keep things equal and so that you could compare the freeformat (standard) source file with the equivalent exclusive source file. Of course, you can't actually do that (for Zero 7 anyway) because there isn't a 384kB CBR available from the exclusive service; 320 being the highest setting. So let's make it easier shall we? Just download the same tracks from their standard AND exclusive service set to LAME standard or extreme and compare the results for yourself.

In our listening tests, we downloaded 384 freeformat source files from a number of albums (Madonna's Ray of Light; Fluke's Puppy; Beatles White Album; George Harrison's All Things Must Pass for example) and decoded them to WAV for burning onto CD-R as you suggest. We then ripped the same tracks to LAME standard/extreme or 192/220 CBR preset from our own CDs and decoded them back to WAV for burning to CD-R. We listened to the freeformat and original PCM sourced files. So in summary, allofmp3's freeformat files were just decoded to WAV prior to audio CD burning. Our own tracks were ripped to mp3 (at supposedly inferior settings) prior to decoding to WAV and then burning. The results were like chalk and cheese, except for tracks from Madonna's American Life, where the freeformat files sounded far closer to ours, our own RIPs were preferred every time; there were some ties on American Life. There are probably many other 384 source files that sound acceptable, but I haven't come across them myself yet.

IMHO, allofmp3 has applied some sort of lo and hi-pass filtering to these freeformat encodings, presumably to cut down on storage space; but to the detriment of overall sound quality. Without the right equipment and know-how, I'm unable to check this; perhaps there's someone out there who can.

Will someone please suggest what the best encoded format is to use is for downloading from alofmp3 ?

Ta

For lossless sourced files, MusePack should be better than MP3 or Ogg Vorbis (as the original Xiph build is used, AFAIK).
For freeformat MP3 sourced files, download AS-IS (384 kbps MP3).

OK thanks. I have downloaded some 384kps mp3s but Windows Media Player refuses to play it. It says it does not recognise the format. Any ideas ? Also do you think 384kps wmp3s would play on an Ipod ?

384 kbps MP3s are freeformat MP3s. Only MAD and some modified versions of mpglib can play back those files. You could try to decode them to WAV using LAME 3.97 (maybe 3.96 supports decoding them, too) and reencode them to whatever format you want, but please note that transcoding will produce an additional quality loss.

OK so what is the next best encode format that will work on Windows Media Player & Ipods ?
320 kbps and Quality "Insane" ?

Well, --alt-preset insane should be the next best (and compatible) setting. However, maybe APS or APE also work good for you (depends on your hearing capabilities).

by the way, here is a list of all albums available as oeex:

http://stat.allofmp3.com/ratings/charts.shtml?action=ll

[Warning: Noob]

So, how did you ascertain that these were the settings used for encoding at allofmp3?

I thought I'd add that, best as I can tell, the site switched over from 3.93. to 3.96 for its online encoding (this happened sometime in the last week of May, according to my encspotting). Of course I don't imagine they would have re-done all their source files.

Basically, I've been trying to find out how how to view encoding information (like the sort you've come up with here) on the 384kbps freeformat files. It'd be nice if I knew how to do it myself (encspot and lametag.exe were a bust).

chrs

yes, they switched to LAME 3.96 but only for their ABR/VBR encodings...for all other CBR files (including their 384kbps files) you order LAME 3.93.1 is still used (with poor settings, using the -k switch and -m s and therefore NOT the presets). They have the settings somewhere on their site...I wouldn't bother with anything else than the presets anyway (abr or vbr modes) and from CD-DA only, because of their poor 384kbps sounding files...I'm also very happy, that they decided to use LAME 3.96 for --alt preset standard , because that's what I'm using

btw: check out this thread: http://www.hydrogenaudio.org/forums/index....=ST&f=1&t=20713 and download the program mentioned there...it will also tell you about the switches made in a mp3. But EncSpot will work as well. I mean, you already see if it is Joint-Stereo or not, you can take a look at the lowpass filter (that will tell you about the -k switch) and everything else is stored in the LAME-Tag as well...

Thanks - you're right, it does say somewhere on the site that they're using those two switches. Wouldn't -k be more responsible for poor sound on their source rips? I'm just thinking stereo redundancies wouldn't eat into quality (much) at those bitrates - hope that doesn't sound naïve.

I ran the program you pointed to - the one thing I was hoping it would do is view lame info on the 384kbps freeformat files (mostly what version was used, of course it would be nice to see stereo and filters too). I wanted to be able to tell when allofmp3 had fixed their archiving habits. But nothing I try seems to 'spot' the freeformat files - do they even have tags?

as I said, they still use LAME 3.93.1 --freeformat -k -b 384 -q 0 -m s. Once more and more people complain they might chance something...but if everybody just sits there and wait nothing will happen...I also thought the these bad settings were responsible for their bad freeformat encodings, but to be sure you have to test it yourself. I haven't done that yet, but just download LAME 3.93.1 and apply the settings I posted above. Then you download one of their freeformat encodings and compare them. Of course, it has to be a CD you have as Lossless and which they offer as 384kbps (they don't offer their lossless stuff as freeformat).

I think it would be very interesting...keep us posted

Thanks again.

They've refunded me for badly encoded albums in the past - and (I notice) removed the albums from their catalogue. So I'd be almost optimistic that a recommendation made on these boards would get a fair hearing.

I guess my next noobie question would be whether --preset insane would be superior to freeformat with ANY switches, because of the code tweaks (also stereo and psychoacoustic redundancies)?

Or just tell them not to use -k and -ms (but stick with 384kbps)?

Anyway, I'm not the expert - but like you said, the more of us that nail our demands to their cathedral door, the better. Just need an 'official line.'

As to other possible reasons for bad-sounding files - I wouldn't rule out that they use the occasional pirated CD, maybe even inadvertently.

Yeah. That is really possible.
I've have one CD (not an group album, but a mix) that is mastered from lossy format (lowpass filter applied) and I bought it at the store !

I think no. They both have a huge bitrate (320vs384 as they use) so there should be no noticable difference at quality.
Potentially, as I understand things, freeformat should be superior, but I think nobody use it...
Gabriel should know it for sure.

You try that! I failed! They just won't do it...even if you explain to them that the -k switch will make songs sounding worse, even though there are many frequencies...they simply don't care


yes, this could be true. They also accept songs from their community and pay for that. Just imagine some newbe messing around with LAME switches

Anyway, someone in this thread mentioned that a album that was available as CD-DA sounded bad as well when not taken from the original source...but the CD-DA source was fine...so that would mean that they mess it up...

It just came in my mind:
Also, the service POSSIBLY can be based on someone's personal huge collection, which was encoded, for example in layer 2.
And they took it and reencode to freeformat mp3 thinking it will not degrade in quality...
There can be very many options, why this happened.
I hope they could learn on their mistakes and situation will improve...

Guess we need someone who speaks Russian but I will try sending them a link to this thread...

In their favour, compare allofmp3 to the other players in the Russian market - ie. the mp3key.net sites, whose business model really doesn't differ from the average ringtones site.

I think allofmp3's intentions are fairly good, and they've made their operation pretty transparent. Yes, their transcoding would be better if they archived in lossless - but I'm sure their resources are fairly limited, and server space is an issue for them (remember how hard it used to be accessing their site). Plus transcoding from lossless takes longer, and most shoppers are (let's face it) more interested in convenience than quality.

In short, designing a by-request transcoding service, where time and storage space are both restricted, is a bit of a balancing act, and allofmp3 probably believe that they have found the best solution. Just gotta keep hammering away at them till they realise there's a better one.

Indeed!


who says that?! It should take the same time as from a lossy source! I think they cache their downloads anyway...so if someone encodes a song they probably keep that files for some days until no one encodes it during a certain amount of time.

@ SirGrey

I doubt it. Have you read this thread here entirely? Some people claim that their files from CD-DA are fine, but when you order a file, which is available as CD-DA, but choose the cheaper method and download the 320kbps file and it sounds bad...