HydrogenAudio

Hello All

I have not been keeping with the latest goings on in the last few years.

Is these a consensus on an audio format vs file size that is mostly transparent?

I have always used V2 with LAME. (I still remember the --r3mix option!)

Currently I encoded a bunch of stuff with -V4 for portable usage and it has been great.

Is there any other file formats or options that give the smallest file size that is mostly transparent?

I know mostly transparent is not a scientific analysis but I am unsure if there is a happy medium now?

Thanks in advance.

For portable use, probably -V5 in lame would be fine.

You could also look into lower bitrate AAC or Vorbis.  When I still used my ipod, I'd often use 100k VBR on it, and was quite happy with the quality.

For portable use, probably -V5 in lame would be fine.

You could also look into lower bitrate AAC or Vorbis.  When I still used my ipod, I'd often use 100k VBR on it, and was quite happy with the quality.

Yeah thanks.  So no testing or threads about?

100k VBR is probably getting a bit low. But I also do not know how much quality at the lower end has improved.

I found some stuff that was encoded with Xing and Blade from 1999 on my computer. My ears are still bleeding!

There is a search function in the top-right corner. I bet You will find a lot of useful information in this forum.

For transparency and portability, it's still tough to beat LAME and MP3.  I find -V4.5 -q0 (~160kbps) to be fine for portable players, though to be on the safer side I usually -V0 everything.
You may also want to check out halb27's LAME variant.  It increases the accuracy requirements for tough-to-encode sections.  Of course that can dramatically raise the bitrate too.

There is a search function in the top-right corner. I bet You will find a lot of useful information in this forum.

Lots of is the key. Nothing specific. Hence the new thread. ")

I can even suggest you AAC True VBR q36 (~95kbps) but you have to ABX to better know what you need.

Quote from: saratoga on 2012-11-13 02:42:09

For portable use, probably -V5 in lame would be fine.

You could also look into lower bitrate AAC or Vorbis.  When I still used my ipod, I'd often use 100k VBR on it, and was quite happy with the quality.

Yeah thanks.  So no testing or threads about?

100k VBR is probably getting a bit low. But I also do not know how much quality at the lower end has improved.

I found some stuff that was encoded with Xing and Blade from 1999 on my computer. My ears are still bleeding!

If you just want over 100k and don't care too much about perfect transparency you could probably use virtually any modern codec.

Check hardware specs of the target player and if it supports AAC or OGG there's no good reason (except slightly lesser power consumption) to choice mp3. From all modern lossy codecs it has worst quality/bitrate ratio. I use for my phone AAC at ~192k-VBR which is kind of overkill but the advantage is the same files can be shared between portable, PC or interior multimedia player where they still sound transparent enough.

Is this for keeping your only copy, or is it for transcoding from your lossless archive to get maximum minutes on a portable device with limited storage capacity? In the latter case, you can of course start low and then reencode if you need it.

Why don't take a different approach to this matter at hand, bitrate vs transparency tradeoff: instead of searching an abstract value of bitrate mostly transparent, let's start from the real storage available and calculate how much music we could expect to put in it.

I explain: I have loaded my DAP with 8Gb of 256kbps VBR AAC, completely transparent to me, reaching a grand total of about 70 CDs, sometimes more sometimes less, so it's about 9 CD per Gb. @160kbps, a bitrate "mostly transparent" but which I successfully ABXed more than once with harpsichord (which is an instrument I happen to like a lot!), I would have gained 30, maybe 40 more CDs, but then I would have been listening in constant fear of compression artifacts: I decided for peace of mind!
 
And is a fact that nowadays DAPs have reached a quality level so that they can be considered a primary source, using good headphones or options like USB, line or digital out, AirPlay etc...

Is this for keeping your only copy, or is it for transcoding from your lossless archive to get maximum minutes on a portable device with limited storage capacity? In the latter case, you can of course start low and then reencode if you need it.

Transcoding. I have my FLAC collection.

I get the just of advice is to try. I guess there is no medium any more.

Not really looking for the "I can compress more, so I can fit more" It's more of a case of "this is about transparent for the lowest file size."

IMO, MP3 takes the crown .  Others may be better towards the 120k side of thing but for a long time now its no big deal at all to go V4 ~ 153 kbit . Quality is almost always very good if not perfect not to mention the compatibility.

Opus at ~128 kbps is also pretty good (I couldn't ABX it on my usual test track last time I tried). Rockbox dev builds support it, gonemad on Android partially supports it, Poweramp on Android will support it in the next (or so) release.

I hope the developer on DeaDBeeF changes his mind about his opus priority. For portable use in noisy or almost noisy environments I could probably go as low as 96kbps using opus.

IMO, MP3 takes the crown .  Others may be better towards the 120k side of thing but for a long time now its no big deal at all to go V4 ~ 153 kbit . Quality is almost always very good if not perfect not to mention the compatibility.

Yeah, I am thinking of staying with LAME and -V4.

Test folder of songs.

793 tracks and average bitrate was 151kbps.

I wanted to test QAAC but it is just too difficult to setup. No simple guide on how to get it to work with dBpoweramp or Foobar.

I am currently using a Galaxy Nexus 16gb and due to the Nexus 4's lack of space I'll be stuck with 16gb again, so smallest file size without compromising quality too much.

Based on my own limited testing, I'd say the lowest bitrate transparency for music can be achieved with AAC. More specifically with either the FhG (Winamp) or Apple's encoder at around 130 kbps.
The Winamp's AAC encoder can be used with Foobar as well (there's a command line encoder somewhere).

At higher bitrates (170-180) they all sound pretty much transparent to me, including MP3.

Sure, you can still ABX some even at 200+ with careful listening, but for casual listening outside of ABX, I'd have no problem recommending something like FhG at setting 4 (~130kbps).

Do some test encodes of OGG Vorbis or AAC targeting around 96kbps and try them.. I'm betting you'll be shocked how good they are..
Once Opus has better player support it would probably get the nod from me, but right now I'd say AAC and Vorbis are the best options for what you want. It seems silly to use mp3 at all unless your player supports nothing else.

A personal listening test of AAC and MP3. http://d.hatena.ne.jp/kamedo2/20111029/1319840519 (http://d.hatena.ne.jp/kamedo2/20111029/1319840519)
I won't be tired to mention kamedo2's  tests again and again. He is a great listener.

So the graph shows that Apple AAC 96-100 kbps is equivalent to LAME MP3 160 kbps.

Hydrogenaudio Listening Tests (http://wiki.hydrogenaudio.org/index.php?title=Hydrogenaudio_Listening_Tests)

There is an easy way to make a portable  version of QAAC encoder without installing Apple Quicktime or iTunes.
Then it's simple to use QAAC as any other command line encoder with foobar converter.
http://www.hydrogenaudio.org/forums/index....st&p=785744 (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=85135&view=findpost&p=785744)

OK Thanks IgorC  I did take a look at the listening tests and took from it that QAAC was the best.

Just frustrating there is not a simple plugin built for foobar, or insutructions that are clear and easy to find.

I understand the legalities etc though.

I'll give it a shot and see how I go around 96-100kbps.

Thanks.

Just a thought, for less than $40 you can get a DAP that takes 32 GB micro SD cards. I often fill one with FLAC files instead of MP3 because I don't really need several thousand files at a time.

Yeah I could get another MP3 player, but I am only interested in using my phone now.
I also do not like to have to change stuff all the time. I want to put a couple of thousand songs.

So is there a guide to QAAC? It is so confusing. So many options and incompleted quides.

Why is there nothing like this for AAC?

http://wiki.hydrogenaudio.org/index.php?ti...ncoder_settings (http://wiki.hydrogenaudio.org/index.php?title=LAME#Recommended_encoder_settings)

Just makes it hard trying to get it working with Foobar2000 when there is so clear guide on recommended options etc.

I cannot even find a guide that gives examples of Foobar command lines for QAAC.

So I have managed to get this to work:

qaac.exe in Foobar2000.

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

If so I'll play around that bitrate area.

Thanks everyone

For the FhG AAC encoder see here (http://www.hydrogenaudio.org/forums/index.php?showtopic=97687). The encoder .exe is here (https://github.com/tmkk/fhgaacenc/downloads) and the other files you get from Winamp.
Try --vbr 3 and --vbr 4. The latter should give you very good results.

check QAAC's wiki, https://github.com/nu774/qaac/wiki (https://github.com/nu774/qaac/wiki)
it has a page: "Command Line Options".

Is there a somewhat consensus that FhG is better than QAAC now?

I note the listening tests are about a year old now.

Thanks

check QAAC's wiki, https://github.com/nu774/qaac/wiki (https://github.com/nu774/qaac/wiki)
it has a page: "Command Line Options".

I did see that, but it was not clear. Are there any addition options needed for highest quality? or has it been simplified to just the bitrate?

There was also not many Foobar examples.

I do not mean to criticise at all. Just hard trying to figure it all out and even getting the encoder files are hard enough!

my foobar2000 command line for approximately 128 kbps (for example):
--ignorelength --no-optimize --quality 2 --silent --tvbr 63 - -o %d

for very high vbr (around 320 kbps, or more) just put 127 instead of 63.

a command line for about 96 kbps would be:
--ignorelength --no-optimize --quality 2 --silent --tvbr 45 - -o %d

the "--quality 2" is default (and the highest), but I still use it. "--no-optimize" is probably useless to anyone other than me - it just doesn't optimize the mp4 container.

there is no "-q 0" or "-q 9" like in LAME; there is "--quality 0", but that means less quality.

there's less options related to quality when it comes to qaac, and maybe all AAC encoders.  Apple has the best encoder according to listening tests, which is what qaac uses anyway.

almost forgot about "--cvbr".  constrained VBR allows for more control, like:
"--cvbr 96" would give a little bit more controlled bitrate at about 96 kbps, but it's still vbr.

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

Yes, it's the highest quality setting for true VBR (-V, TVBR ). You can try a constrained VBR as well (-v, CVBR) -v 96 for comparable bitrate. CVBR has  higher score than TVBR though without statistic difference. http://listening-tests.hydrogenaudio.org/i...-96-a/index.htm (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/index.htm)

Is there a somewhat consensus that FhG is better than QAAC now?

I note the listening tests are about a year old now.

Thanks

The situation hasn't changed since then. There were some small tunings but major releases remain the same.
AAC encoders were tested since a long time ago http://en.wikipedia.org/wiki/Codec_listening_test (http://en.wikipedia.org/wiki/Codec_listening_test)
And Apple encoder is/was always on top. It could be the case that Apple LC-AAC encoder has reached the best possible quality for the entire AAC format. Possibly there are some other LC-AAC encoders with same or near the same quality but never better than it.

Is there a somewhat consensus that FhG is better than QAAC now?

I note the listening tests are about a year old now.

Thanks

I don't think there's such a consensus, no.

I did like FhG a tiny bit better than QAAC on a single sample I compared them with, but that's definitely not enough to draw any conclusions. Besides one might perform better than the other only at specific bitrates.
I prefer to use it simply because it's a bit easier to use with Foobar, technically (perhaps not legally) you don't need to install anything, just extract a couple of files.

my foobar2000 command line for approximately 128 kbps (for example):
--ignorelength --no-optimize --quality 2 --silent --tvbr 63 - -o %d

for very high vbr (around 320 kbps, or more) just put 127 instead of 63.

a command line for about 96 kbps would be:
--ignorelength --no-optimize --quality 2 --silent --tvbr 45 - -o %d

the "--quality 2" is default (and the highest), but I still use it. "--no-optimize" is probably useless to anyone other than me - it just doesn't optimize the mp4 container.

there is no "-q 0" or "-q 9" like in LAME; there is "--quality 0", but that means less quality.

there's less options related to quality when it comes to qaac, and maybe all AAC encoders.  Apple has the best encoder according to listening tests, which is what qaac uses anyway.

almost forgot about "--cvbr".  constrained VBR allows for more control, like:
"--cvbr 96" would give a little bit more controlled bitrate at about 96 kbps, but it's still vbr.

THANK YOU! Thank explains it! Makes perfect sense. I am encoding a bunch of 20 tracks at 96 kbps and see how it sounds. (Yes, I know that is not scientific etc)

Quote from: funkyblue on 2012-11-13 15:13:43

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

Yes, it's the highest quality setting for true VBR (-V, TVBR ). You can try a constrained VBR as well (-v, CVBR) -v 96 for comparable bitrate. CVBR has  higher score than TVBR though without statistic difference. http://listening-tests.hydrogenaudio.org/i...-96-a/index.htm (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/index.htm)

Quote from: funkyblue on 2012-11-13 15:21:34

Is there a somewhat consensus that FhG is better than QAAC now?

I note the listening tests are about a year old now.

Thanks

The situation hasn't changed since then. There were some small tunings but major releases remain the same.
AAC encoders were tested since a long time ago http://en.wikipedia.org/wiki/Codec_listening_test (http://en.wikipedia.org/wiki/Codec_listening_test)
And Apple encoder is/was always on top. It could be the case that Apple AAC encoder has reached the best possible quality for the entire AAC format. Possibly there are some other AAC encoders with same or near the same quality but never better than it.

Thank you as well. I'll stick with QAAC now it is working.

Thanks again everyone. Much appreciated.

I am truly shocked at the quality of 96kbits!

While I can hear some artifacts, it is still very good.

I'll do some abxing on 112 tomorrow.

Things have really come along way from the 128kbit Blade days

What kind of phone? All Android phones will play OGG Vorbis and Foobar2000 will natively encode to it as far as I know.

I am currently using a Galaxy Nexus 16gb and due to the Nexus 4's lack of space I'll be stuck with 16gb again, so smallest file size without compromising quality too much.

So I have managed to get this to work:

qaac.exe in Foobar2000.

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

If so I'll play around that bitrate area.

Thanks everyone

Q0 - Q4 (0) = ~40 Kbps
Q5 - Q13 (9) = ~45 Kbps
Q14 - Q22 (18) = ~75 Kbps
Q23 - Q31 (27) = ~80 Kbps
Q32 - Q40 (36) = ~95 Kbps
Q41 - Q49 (45) = ~105 Kbps
Q50 - Q58 (54) = ~115 Kbps
Q59 - Q68 (63) = ~135 Kbps
Q69 - Q77 (73) = ~150 Kbps
Q78 - Q86 (82) = ~165 Kbps
Q87 - Q95 (91) = ~195 Kbps
Q96 - Q104 (100) = ~225 Kbps
Q105 - Q113 (109) = ~255 Kbps
Q114 - Q122 (118) = ~285 Kbps
Q123 - Q127 (127) = ~320 Kbps

I could ABX 112kbits. Could not ABX files in Lame -V4 3.100 Alpha.

I'll give 128kbits a shot next.

I could ABX 112kbits. Could not ABX files in Lame -V4 3.100 Alpha.

And no, a separate ABX tests aren't useful to compare different codecs.  Prejudice will always be there.
One should perform blind test with two encoders at the same time (ABC/HR).

It appears that funkyblue is only using ABX to discover the level of transparency for each codec, and will then select the codec whose transparency level uses the least bits. This is a perfectly valid use of ABX.

If he were comparing codecs that are not quite transparent to choose the least offensive one then ABC/HR would be appropriate.

It's not that simple and linear.

The conditions between two ABX sessions vary a lot. Add to it  personal prejudice about that higher bitrate produce better results (no matter what encoder) and faster abort of ABX session with claim of transparency.  Typical case.

Shortly ABX isn't meant for comparison between two and more encoders. Well, If You search for average level of transparency but later still You  will try to make a conclusion that encoder A is better than encoder B . So, no go. Only ABC/HR.

Is there a somewhat consensus that FhG is better than QAAC now?

I note the listening tests are about a year old now.

Thanks

FhG = Fraunhofer and it's ok to say it because they make the AAC codec but QAAC doesn't mean anything, QAAC uses Apple libraries to create AAC so it's Apple.

Apple still better.

But wasn't the only AAC comparison done at ~96kbps? Isn't that insufficient to say one codec is better in general?

AAC achieve transparency at around 128kbps, obviously they are all "good" over ~128kbps and if one of them is better at ~96kbps it means it's better in general.

Now, a different discussion can be made for HE and HEv2 but I don't use them, I don't know enough.

It's not that simple and linear.

The conditions between two ABX sessions vary a lot. Add to it  personal prejudice about that higher bitrate produce better results (no matter what encoder) and faster abort of ABX session with claim of transparency.  Typical case.

Shortly ABX isn't meant for comparison between two and more encoders. Well, If You search for average level of transparency but later still You  will try to make a conclusion that encoder A is better than encoder B . So, no go. Only ABC/HR.

That is true. I lean towards LAME -V4 because it is larger in size. My mind still has issues with low bit rates.

That being said I just tried to ABX 5 songs at random as a quick test on 112 AAC and could not.

Is there a guide to setup and do a proper test for say 10 tracks and I can pick various bitrates?

It is annoying in Foobar to have to do 2 at a time and listening to the same samples over and over.

I'm glad You are open mind. 

Here is a guide
http://www.rarewares.org/rja/ListeningTest.pdf (http://www.rarewares.org/rja/ListeningTest.pdf)

ABC/HR application
http://www.hydrogenaudio.org/forums/index....st&p=683924 (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=77573&view=findpost&p=683924)

Quote from: funkyblue on 2012-11-13 15:13:43

So I have managed to get this to work:

qaac.exe in Foobar2000.

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

If so I'll play around that bitrate area.

Thanks everyone

Q0 - Q4 (0) = ~40 Kbps
Q5 - Q13 (9) = ~45 Kbps
Q14 - Q22 (18) = ~75 Kbps
Q23 - Q31 (27) = ~80 Kbps
Q32 - Q40 (36) = ~95 Kbps
Q41 - Q49 (45) = ~105 Kbps
Q50 - Q58 (54) = ~115 Kbps
Q59 - Q68 (63) = ~135 Kbps
Q69 - Q77 (73) = ~150 Kbps
Q78 - Q86 (82) = ~165 Kbps
Q87 - Q95 (91) = ~195 Kbps
Q96 - Q104 (100) = ~225 Kbps
Q105 - Q113 (109) = ~255 Kbps
Q114 - Q122 (118) = ~285 Kbps
Q123 - Q127 (127) = ~320 Kbps

Are the numbers next to the Q ie (82) the middle value?

Quote from: eahm on 2012-11-13 18:02:18

Quote from: funkyblue on 2012-11-13 15:13:43

So I have managed to get this to work:

qaac.exe in Foobar2000.

-V45 -o %d -

Average of 96kbits. Is that the highest quality option for that bitrate?

If so I'll play around that bitrate area.

Thanks everyone

Q0 - Q4 (0) = ~40 Kbps
Q5 - Q13 (9) = ~45 Kbps
Q14 - Q22 (18) = ~75 Kbps
Q23 - Q31 (27) = ~80 Kbps
Q32 - Q40 (36) = ~95 Kbps
Q41 - Q49 (45) = ~105 Kbps
Q50 - Q58 (54) = ~115 Kbps
Q59 - Q68 (63) = ~135 Kbps
Q69 - Q77 (73) = ~150 Kbps
Q78 - Q86 (82) = ~165 Kbps
Q87 - Q95 (91) = ~195 Kbps
Q96 - Q104 (100) = ~225 Kbps
Q105 - Q113 (109) = ~255 Kbps
Q114 - Q122 (118) = ~285 Kbps
Q123 - Q127 (127) = ~320 Kbps

Are the numbers next to the Q ie (82) the middle value?

Those are the numbers qaac pushes when you select between the min and the max. For example -V79, -V80... -V85, -V86 they all give -V82. I think it's a True VBR average, that maybe nu774 or Apple set up, I don't know precisely who and I didn't test qtaacenc to see if the result is the same.
I've actually put them there myself, testing qaac, for an easy reminder of the exact setting I have to insert when I save on foobar2000.