I wish there's a general section under lossy audio compression. Haven't encoded in lossy for ages, the good old days was back in lame 3.97...
Anyway, would like to encode Andrea Bocelli - Opera (2012) in the smallest file size possible that can be handled by blind ABX! I'm open to any lossy encoder that's best for the job. I also wonder what info I have to provide about this CD to get the best conclusion here.
cheers.
I wish there's a general section under lossy audio compression.
This one that I moved it to works as that.
Anyway, would like to encode Andrea Bocelli - Opera (2012) in the smallest file size possible that can be handled by blind ABX! I'm open to any lossy encoder that's best for the job. I also wonder what info I have to provide about this CD to get the best conclusion here.
Do you understand the purpose of ABX testing? It’s totally specific to an individual user, hardware, and sample. How can you expect other people to be able to tell you what you will or will not hear?
The most you’ll get, quite likely, are very general statements like these: AAC/MP4 is good, MP3 is good, Vorbis is good, Opus is getting good, and decide for yourself between these based upon their features, your preferences, and the results of tests that you conduct yourself.
I don’t mean to dissuade anyone who might have more specific advice, but I don’t know what other people can say in response to the request for information on ABXing, which is by definition not something that others can do on your behalf.
You will be inundated with replies to "test for yourself" however I am happy to give you my experience from tons of my own testing. Vorbis -q2 has given me that magic 10% compression threshold that i can't abx. Fhg aac I could _barely_ abx at q3 can't do it at q4. I've heard Nero or Apple aac is somewhat better than fhg but I haven't bothered to test. Mp3 and WMA standard need higher filesizes for transparency. The nice thing with vorbis is if you detect anything you can go up in smaller increments, ie, 2.1,2.2 until you reach that point of transparency for you on your equipment. WMA pro is good too but limited in its hardware support. But I've been 100% satisfied with libvorbis -q2.
Come on db1989, the guy is just looking for a starting point, is it so wrong for you to say "based on my experience I use x codec with x settings for my equipment? I showed mine, let's see yours XD. With all the codecs out there its a jungle.
Nowadays I very rarely convert or purchase new music in lossy formats and tend to simply settle for MP3 at 192 kbps or higher (VBR if I can get it) without worrying about testing because I’m careless like that. How does this personal anecdote help someone who is specifically asking for a ready-made setup that will be transparent to them?
Not that more specific information about tests, if I had done any, would be any more applicable, really. The question cannot be answered in its specific form in a way that is guaranteed to be reliable, and general advice will probably just end up mirroring very broad sets of consensuses arrived at by previous tests and threads, all of which are available by searching. You’ll excuse me if I state this up-front to the OP rather than implying that personal accounts from other users can be relied upon to be applicable to him/her.
And anyway, as I said, other users are free to post their own impressions and habits once these caveats have been made clear. I simply chose not to describe mine for the reasons that I already explained and because my methodology for choosing settings is almost nonexistent and not something that I would recommend ‘officially’. There’s no need to imply that I’m withholding information out of some Scrooge complex.
I know but it just seems to be a tendancy around here to hammer new users with "test for yourself". Often times I'm curious myself what codec/settings experienced folks like you or greynol or Arny use even though I've done way too much tinkering and abx testing myself. Sometimes people just like a starting point, ya know?
As a good starting point, you could start at 96Kbps with the following codecs:
Opus
LC-AAC: Apple AAC or fdk-aac or Nero AAC
Vorbis
Then just increment the bitrate each time until you don't hear any audible differences on one of the codecs against the source.
LithosZA, remember fdk-aac (https://github.com/mstorsjo/fdk-aac) is the open souce AAC Encoder from Fraunhofer BUT fhgaacenc (https://github.com/tmkk/fhgaacenc) is a CLI for the official Fraunhofer AAC Encoder. fhgaacenc requires Winamp libraries.
I would start testing from 96 kbps like LithosZA said but I would give priority to the Apple AAC Encoder (qaac (https://sites.google.com/site/qaacpage/) or qtaacenc (http://tmkk.undo.jp/qtaacenc/)) since it has been proven (in 2011 (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/results.html)) that it's the best AAC Encoder available. qaac requires Apple libraries from iTunes or QuickTime. qtaacenc requires QuickTime to be installed.
I would maybe test MP3 (LAME (http://www.rarewares.org/mp3-lame-bundle.php) of course), starting from 128 kbps and again maybe Vorbis (aoTuV (http://www.rarewares.org/ogg-oggenc.php)) starting from 96 kbps. Since I switched to ACC I no longer like MP3, I think it's old technology and should disappear (not going to happen). I like Vorbis for its quality and because it's open souce but it's not compatible with the main portable device I use which is an Apple device.
I would skip Nero AAC (good quality but barely developer anymore). I would skip Opus as well since I don't think it's ready yet (many will disagree).
I know but it just seems to be a tendancy around here to hammer new users with "test for yourself".
How a very specific question like this from the OP:
would like to encode Andrea Bocelli - Opera (2012) in the smallest file size possible that can be handled by blind ABX!
could possibly be answered differently? We don't even know his age!
In general terms I can answer that I have my whole CD collection, about 900 classical CDs, some of them vocal, encoded with Apple AAC, VBR at quality level 110 and haven't find a single transparency issue till now.
But, as I think most of us, I haven't done extensive ABX tests for each and every track, as I actually like to listen to music, not to test codecs!
That said, it is likely that some programmes like operatic stuff could be encoded at a lower quality level still resulting transparent. A sensible advice could be to start with a mid quality setting for the codec of choice, find some artifacts and then work on them raising the setting by little increments.
And if one wants the lowest possible file size, must do it for every track of that album: he'll end up to hate it!
Edit: given the premise, a serious valid advice is to use a modern codec in VBR mode, target for
quality, thus forgetting kbps, and let the codec does his job which is to find the lowest bitrate for that desired quality.
By the way: HAPPY EASTER TO EVERYONE, MEMBERS AND LURKERS!!!
@Mach-X, thanks for understanding : )
@db1989, thanks for your opinion, I guess I understand the purpose of ABX. But I was actually requested by a friend for the lossy encoding. I just think there's some member here who has the knowledge and 'adequate' hardware and experience could give me a starting point.
Bottom line, too much ABX test will drive me crazy!
If too much abx drives you crazy, just start for a moderate mp3 setting like V4. That setting actually works really well & better quality than the popular portable setting like V5.
OK, someone wants a copy in a lossy format, and you do not wish a "sounds crap!" complaint. Use mp3 for max compatibility and LAME at the 0 setting and save yourself time.
But if the person does not have the original to compare to ... ? Heck, send him/her also an AAC at 96 (two-figure numbers are more impressive!) Then (s)he can compare the mp3 to the AAC, and if (s)he uses a visual tool (spectrum analyzer or whatever), (s)he could even pick the wrong one by comparing apples and oranges ;-)
So far, Mach-X has given a great starting point, Vorbis -q2
As a sanity test, I encoded a Motorpsycho song (long-haired, bearded epic rock) to V5 with lame. This also downsampled it to 32000Hz.
I was unable to instantly ABX that from the original, and didn't feel like trying too hard. I usually give up on ABX if the difference isn't night and day.
That said, it's growly rock, and there wasn't much real content above 16KHz anyway, but still it's interesting to note that depending on the properties of the music, you can go really low with settings and still not mess up a song.
You'll have to try for yourself* to see if that's also true for your opera song. Just pick a ridiculously low quality setting at first and see how you fare.
*) The reason we keep saying this, Mach-X, is because it's true.
@Mach-X, thanks for understanding : )
@db1989, thanks for your opinion, I guess I understand the purpose of ABX. But I was actually requested by a friend for the lossy encoding. I just think there's some member here who has the knowledge and 'adequate' hardware and experience could give me a starting point.
Bottom line, too much ABX test will drive me crazy!
As a rule of thumb, most codecs do OK with acoustic instruments but can struggle with artificial sounds, so classical doesn't often require excessive bitrates and doesn't have many killer samples remaining (except things like harpsichord). Often, VBR settings can come out a little lower in bitrate for classical than for pop/rock.
If it's for a friend, sticking with widely supported formats might be best so they can just play it anywhere without new software or codec packs, though if they don't use specific class of software or hardware but get one file per track, gapless playback may be the first casualty (if tracks have any continuous sound between them - as a live concert might). Then again, even in live classical, track transitions are usually within a near-silent moment after applause has died down.
For maximum compatibility even on old portable players and cheaper digital TVs with USB thumb drive music playback, MP3 (e.g. LAME at -V3 or -V2 should be transparent for easy-to-encode stuff like real instruments) - 170 to 200 kbps range. -V5 (about 130 kbps) or -V4 (about 150 kbps) are rarely annoying, albeit that differences might be detectable occasionally under critical headphone listening if you know what lossy artifacts sound like.
For pretty widespread compatibility in 2013, AAC-LC in a .m4a file format is good and the Apple Quicktime/iTunes encoder is probably consistently highest rated (and can be accessed using QAAC), and is probably fractionally more consistent in standard bitrate-choice mode (often called CBR, but it's actually CVBR). 96 kbps often sounds very good, rarely annoying, albeit that differences might be detectable occasionally under critical headphone listening if you know what lossy artifacts sound like. 128 kbps is very close to transparency most of the time and would be a fine choice. A lot of people go a little higher in bitrate.
I'd be tempted to use LAME -V4 (about 160 kbps) or QAAC/iTunes/Quicktime 128 kbps to offer very good quality on the verge of transparency along with modest file size and wide compatibility, though QAAC at 96 kbps is also tempting, as is LAME -V5 (~130kbps), especially if I'm transmitting over the internet with modest upload speed.
Public listening tests from the last 5 years or so can be a fairly good guide once you have an idea of compatibility issues. Wikipedia lists a number of public and personal codec listening tests (http://en.wikipedia.org/wiki/Codec_listening_test). The results (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/results.html) of 2011's 96 kbps LC-AAC test (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/index.htm) are quite good over some pretty demanding samples under close scrutiny and Apple and FhG encoders both score well.
You can try QAAC with highest setting in True VBR, and see what the size it gives.
Cause it should be Transparent, and give a nice adaptive bitrate.
Dynamic, everything you said is in my post. Did you even bother reading previous posts?
Also, CBR is CBR, and VBR is not often called CVBR but Apple's VBR is CVBR. Apple don't ever use CBR, they stick with ABR, they would actually use CVBR but the choice of ABR is because it gives a closer bitrate to what people want to see.
zerowalker, -V127 is a little too much IMO.
That "proof" about Apple being the best is hardly proof, especially when it shows that Apple is statistically tied with fhg.
Dynamic, everything you said is in my post. Did you even bother reading previous posts?
Please try to resist the temptation to act offended if people repeat some things that you personally have said, with the implication that they’re not allowed to phrase some of the same things in a different way or perhaps are doing so deliberately as an insult to you. Without meaning anything negative about your post, Dynamic’s post is not a carbon copy of it, and people might appreciate the extra information or better understand repeated things based upon different methods of phrasing (in either post).
Often times I'm curious myself what codec/settings experienced folks like you or greynol or Arny use even though I've done way too much tinkering and abx testing myself.
Lame 3.98.4 -V3 with no special switches except the one telling the encoder not to waste any time calculating RG information that barely has any support. Oh wait that isn't true. I use the scale command in an attempt to achieve equal album loudness also.
Does this answer the OP's question as it was asked? No, I'm sure I could get smaller files that give me the same peace of mind about quality using AAC. There is no other lossy format I would consider because of compatibility reasons.
Does pointing him to performing ABX tests for himself the best answer to the original post? Yes, until we found out it was for a friend.
(long-haired, bearded epic rock)
I'd definitely be able to ABX their beards from Bocelli's yes, at least on certain occations.
OT
Bocelli lives 30mins from where I am from in Italy. We went to see him in Vegas (MGM) last year and it was amazing. This year he will be where I live at the moment (Phoenix, AZ) for Christmas.
It’s not as interesting an anecdote as yours, but you reminded me that my mum went through a total Bocelli phase back in the day.
The FAQ, wiki entries (see links in the old sticky threads for each codec), and usual advice about codec and bitrate choice are as relevant to opera as they are to anything else.
Mp3 and FLAC are popular here for good reason.
Cheers,
David.
Does this Andrea Bocelli recording have special properties which would make it specifically suitable/unsuitable for a certain encoder?
In my mind that is kind of a trick question because I think of lame mp3, apple aac et al. as general purpose tools. Considering e.g. harpsichords though, certain encoders might have an edge.
Does this Andrea Bocelli recording have special properties which would make it specifically suitable/unsuitable for a certain encoder?
In my mind that is kind of a trick question because I think of lame mp3, apple aac et al. as general purpose tools. Considering e.g. harpsichords though, certain encoders might have an edge.
The OP can answer for him-/herself, but it could just be due to the OP having observed that this forum seems to appreciate more information and frown at general uninformative thread titles like "help me choose codec" – in which case I think it is perfectly OK to let the forum actually explain whether this piece of information does matter or not.
I remember the digital album from itunes, "Andrea Bocelli – Notte illuminata" was absolute horrible, a total waste of money and 'talent', IMO! If I might add, "WTF!" Out of respect for the 'professionalism' of Apple, I have to wonder what went wrong! Is it because of AAC mp4a or Apple, or both!?
So I asked this question to see what codec people use for tenor? "Test it yourself" is probably the best way to avoid arguing, what's good for oneself may not for other. Anyway, I do appreciate others' opinions. Even one can provide the ultimate track for testing, this kind of question will be asked from time to time, because technologies evolve, I certainly hope to keep it that way, don't u?
I'm re-'testing' the whole album with my old android phone now, encoded by oggenc2 -q 2. So far, I'm happy with the sound and the size.
cheers.
I have to wonder what went wrong! Is it because of AAC mp4a or Apple, or both!?
Until you have the original source material you will never know. Otherwise this anecdote is completely useless, especially in light of this discussion and your stated goal.
EDIT:
Please read the next post by IgorC. It pretty well summarizes how I feel about this post I quoted as well as the sentiments discouraging ABX in favor of a monkey-see-monkey-do philosophy, both as a member and as a moderator.
I remember the digital album from itunes, "Andrea Bocelli – Notte illuminata" was absolute horrible, a total waste of money and 'talent', IMO! If I might add, "WTF!" Out of respect for the 'professionalism' of Apple, I have to wonder what went wrong! Is it because of AAC mp4a or Apple, or both!?
Sorry, do you realize how is it disrespectful for all these people who were kind with you and
actually have answer your questions and have provided solid information (that we have worked together and for free!) while you demonstrate with your pure anecdotical statements that you haven't take the time to understand their answers or simply don't care.
I remember the digital album from itunes, "Andrea Bocelli – Notte illuminata" was absolute horrible, a total waste of money and 'talent', IMO! If I might add, "WTF!" Out of respect for the 'professionalism' of Apple, I have to wonder what went wrong! Is it because of AAC mp4a or Apple, or both!?
Sorry, do you realize how is it disrespectful for all these people who were kind with you and actually have answer your questions and have provided solid information (that we have worked together and for free!) while you demonstrate with your pure anecdotical statements that you haven't take the time to understand their answers or simply don't care.
What the heck are u on about? How am I being disrespectful as I merely stated the incident with Apple? And how u claimed in the group as "we" but offer nothing in the discussion. So what have u provided as 'we'? And if not free, how much do u want to charge? Have u actually read the posts, or just spam on any one word u don't like?
@greynol, so what's this monkey-see-monkey-do philosophy? Are u denying there's a best way to do certain thing? Even this can change in time!
And for what I've said, of course I blame Apple, and what's that got to do with disrespectful???
I gave a right way with the wrong way; twice now. Also I told you exactly what was wrong with your useless Apple anecdote. In case you don't understand, let me put it another way: what proof do you have that Apple is to blame?
As far as what IgorC has provided, he has taken the time and effort to organize public listening tests so people have some basis for making recommendations and can point to actual data. Unfortunately, people will misinterpret the data and falsely declare winners or losers when there might not be any, but that's the way this stuff goes. Such a declaration was provided linking to such a test. Did you read it? Did you review the results? If so did you see who organized it? Were you asked to pay anything?
...and to continue on with the theme of showing respect for this forum, it's "you" not "u". Refer to TOS #10.
Sorry, it was never an intention of mine to start a conversation that would have people upset at one another. I guess I was looking to have some back and forth with experienced folks regarding real life useage of encoders (what personal music have they abx'ed, what encoder/settings, what headphones/speakers, what in particular do you listen for), as opposed to simply pointing the OP towards our well known test data, which uses samples not everybody has or listens to. I guess now I see the reason nobody likes to answer the 'what codec/settings' questions.
The answer is and always will be ABX or ABC to find out. Involving yourself in this process is far more useful than simply taking someone's advice especially when it is done without any critical thought.
What's the average bitrate using Vorbis q2 with opera? Now read the part in the original post about the smallest size possible.
What's the average bitrate using Vorbis q2 with opera? Now read the part in the original post about the smallest size possible.
I know you were being 100% rhetorical but you asked, George Orff-Fortune, Empress of the World-O Fortuna clocks in at 86kbps average at vorbisq2, according to rockbox. Quick look at all the tracks puts them all in the 80-86 range.
I thought I was but I was completely wrong in my assumption about how the q system works. I could not readily find the answer in our wiki in a cursory search like I can for Lame, but went ahead and committed my words anyway.
I will probably never forget that lower q means lower bitrate and I thank you for that regardless of how foolish it made me look.
I guess I was looking to have some back and forth with experienced folks regarding real life useage of encoders (what personal music have they abx'ed, what encoder/settings, what headphones/speakers, what in particular do you listen for)
Until 2011 there was a common beleive and recommendation to use Nero as AAC encoder and sometimes Apple. "Oh, Joe who is the exprienced audio guy and he recommended to use Nero".
During 2006-2007 Nero HE-AAC was best solution at 48-64 kbps so people have assumed it's the most optimal for high bitrates as well.
But when AAC encoders were tested at 96 kbps it was a big surprise for many people that Nero came out the last (yes, the last) in quality terms.
And ironically now we're getting back with statements like:
I've heard Nero or Apple aac is somewhat better than fhg but I haven't bothered to test.
So if You have a link to real test, results or something that one could say "OK, I buyt it" then post it. Otherwise it's just another "I have heard...", "I was told", "but the experienced people..." etc.
There are some links to some really real tests:
http://listening-tests.hydrogenaudio.org/i...-96-a/index.htm (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/index.htm)
http://d.hatena.ne.jp/kamedo2/20111029/1319840519 (http://d.hatena.ne.jp/kamedo2/20111029/1319840519)
As You can see You have made a wrong statement about Nero encoder and "experienced folks regarding real life usage of encoders" tells as much as nothing.
as opposed to simply pointing the OP towards our well known test data, which uses samples not everybody has or listens to.
Those samples are of most common styles like pop, rock, techno, classic etc (not some synthetic stuff). So the public tests are actually very representative of real usage. And much more representatives than these most used phrases of "experienced people say", isn't it?
So, an anecdotal experiences is simply NOT argument. They're proved to be wrong in 99.99% of cases, especially true with such conditions as personal prejudice and placebo.
I guess I was looking to have some back and forth with experienced folks regarding real life useage of encoders (what personal music have they abx'ed, what encoder/settings, what headphones/speakers, what in particular do you listen for)
By the by, such curiousities could be somewhat satisfied browsing people's answer in a typical poll thread...
So I asked this question to see what codec people use for tenor?
“any”, more or less. In addition you've had a few specific suggestions on not only codec, but also encoder and settings. Not satisfied yet?
Are u denying there's a best way to do certain thing?
To do “a certain thing” for one specified purpose? For the purpose of transparency for a single person on a single encoding, there is the “good enough to be transparent” way, and if that's your only criterion, then all those will be tied. You have existence but not uniqueness of ““a best way””. If you need lowest possible bitrate, then that particular person has to do the testing him-/herself. (Speaking in totally general terms, for g33x, there may even fail to be “a best way”.)
Even this can change in time!
Yeah ... if your goal is merely transparency, then bringing more codecs and encoders to the table, will expand the group tied for first place. If your goal is also lowest possible bitrate, then that person might have to do the testing over and over again. In principle.
And for what I've said, of course I blame Apple, and what's that got to do with disrespectful???
Buying music you don't like and blaming the record store for having destroyed it? It is likely less disrespectful than what I just decided against writing here.
Well, the reason I said I've heard they are better but haven't bothered to test is because I'm from the school of thought that saying 'statistically in this listening test from year 2xxx it was proven as fact this flavour of AAC is better than that one' is meaningless. HOW much better? If the differences are so minute (and I'm sure they are) that at a 96 average bitrate an average person can't distinguish any meaningful differences without abxing 50 times on one specific sample, then it's NOT better, at least on a practical basis. All I was pointing out was that my threshold for fhg aac was quality level 4, but at 3 it was extremely difficult and required heavy concentration. I also said the same of vorbis q2. Yes I am broadbrushing, but sometimes that's all new users are looking for, is a place to start. And in this case the OP, while he got a little feisty for a bit, was completely satisfied with my recommendations.
Your straw-man argument about organized listening tests and dismissal of the hard work that was done in preparing, participating in and presenting a meaningful result is noted. I feel confident that you're a strong proponent for the advocation of discovery and self-sufficiency.
In the meantime, the OP probably still thinks his baseless comments about Apple were OK.
BTW, I searched for all posts with the term foo_abx authored by you and found none.
While not having listened to the digital album about which the OP has been complaining, I’d like to point out that the placebo effect could have been responsible for the perception of terrible quality and that, even if the latter is valid and verified by other listeners, it is not logical to automatically blame Apple or their chosen lossy format. Doing so is just spreading FUD about things that are unconfirmed and whose reasons, if any, are unclear.
I hope posters here can now move past this subject and the associated outbursts, at least until the provision of evidence for the low quality and a convincing argument for why Apple and/or AAC are to blame. I’m tempted to bin it, but that would cause the loss of a few productive or at least educative details buried among the affected posts, so I’ll leave it here unless the SNR declines further.
Does this Andrea Bocelli recording have special properties which would make it specifically suitable/unsuitable for a certain encoder?
In my mind that is kind of a trick question because I think of lame mp3, apple aac et al. as general purpose tools. Considering e.g. harpsichords though, certain encoders might have an edge.
The OP can answer for him-/herself, but it could just be due to the OP having observed that this forum seems to appreciate more information and frown at general uninformative thread titles like "help me choose codec" – in which case I think it is perfectly OK to let the forum actually explain whether this piece of information does matter or not.
I didn't mean to say that the information was irrelevant; the question wasn't rhetorical. Many of the answers so far seem to disregard that the OP is asking for a very specific album. It appears that either people don't know this album or don't deem this detail relevant. If we knew more about the recording, maybe that information could be exploited e.g.: "encoder X will probably bloat because of excessive HF-content" --> less candidates for a DBT.
This
I've heard Nero or Apple aac is somewhat better than fhg but I haven't bothered to test.
and this
If the differences are so minute (and I'm sure they are) that at a 96 average bitrate an average person can't distinguish any meaningful differences without abxing 50 times on one specific sample, then it's NOT better, at least on a practical basis.
are contradictions.
First You comment that some codec is better than other but later when the discussion changes its direction You start to claim that all codecs are all the same.
You contradict yourself let alone other people.
Ok, enough. If Mach-X doesn't get it by now I don't know that he ever will.
Let's follow dB1989's cue and stick to the business of discussing the codec that provides the lowest acceptable bitrate for the title in question and the genre under which it can be classified.
"I've heard" is exactly that. Doesn't mean I'm in agreement, which is why I issued the second statement. Same as for aotuv vs libvorbis. Splitting hairs. There's no straw man argument if there's no argument. Objective listening tests are a fountain of wonderful information and I've hungrily enjoyed all of them and thank IgorC and others who conducted and partook in them. Otoh not everyone wants or has the means to get that technical or in depth or they just want a starting point, which i and others provided, op is happily using them, therefore this thread has served its purpose. Unless anyone else has an actual encoder suggestion that works better than others with opera (of which i would be curious how it works).
Great, fine, wonderful (not quite, but whatever). Let's get back on-topic.