So I was very skeptical about opus when it sounded like a new OGG Vorbis hype where the codec is both free and crappy quality. I didn't think without SBR it would be able to compete with Nero HE-AAC. So I compared at different bitrates and opus always sounded worse. I decided to ABX at different bitrates, opus failed on 32 and 48 but not 64 where the treble was audibly brighter than Nero.
Ok you got me convinced that opus sounds better at 64 kb/s, why does it suck so bad at other bitrates? Also, why didn't you guys include WMA in your listening test along with opus? WMA has proven to be slightly superior quality to HE-AAC at 64 kb/s so maybe opus still has a competitor.
Also, fix your biased testing methodology and keep the bitrates as close as possible to the original! Jesus. My hypothetical codec produces output bitrates of 192 kb/s when my quality setting intends it to be half as less, I demand it to be nominated as capable of transparency at 16 kb/s.
Your unscientific results makes me keep clear of your website these days.
So I was very skeptical about opus when it sounded like a new OGG Vorbis hype where the codec is both free and crappy quality.
Yeah, you’re right; the many people who think it’s a great and promising new development in audio compression must be delusional.
And Ogg Vorbis “crappy quality” too? Relative to what? It tends to beat MP3.
I decided to ABX at different bitrates
Cool story. Where are your results? This certainly doesn’t belong in Listening Tests without, well, evidence of listening tests.
opus failed on 32 and 48 but not 64 where the treble was audibly brighter than Nero.
Were you ABXing the two lossy codecs against each other rather than against the lossless source?
Your unscientific results makes me keep clear of your website these days.
Given your attitude during this entire post, I don’t think much of value was lost.
Ok you got me convinced that opus sounds better at 64 kb/s, why does it suck so bad at other bitrates?
At very low bitrates HE-AAC v2 uses Parametric Stereo which Opus does not. At those rates Opus uses intensity stereo.
You could try to test both HE-AAC and Opus with mono recordings.
Also, fix your biased testing methodology and keep the bitrates as close as possible to the original!
What original? Do you mean 64Kbps? You know that all those codecs were tested at variable bitrate, right?
Also, why didn't you guys include WMA in your listening test along with opus? WMA has proven to be slightly superior quality to HE-AAC at 64 kb/s ...
Where exactly was this proven? Please dig it out. I've done a MUSHRA test WMA vs. HE-AAC at 64 kbps once, and WMA was inferior. Which confirms the results of e.g. the IRT tests here (newer, WMA 8) (http://tech.ebu.ch/docs/techreview/trev_305-moser.pdf) and here (older, WMA 4) (http://tech.ebu.ch/docs/techreview/trev_283-kozamernik.pdf) and Sebastian's test (WMA 10) (http://listening-tests.hydrogenaudio.org/sebastian/mf-64-1/results.htm).
If by "biased testing methodology" you're referring to testing with VBR, then here (http://www.hydrogenaudio.org/forums/index.php?showtopic=97913)'s some recent discussion of that issue. I also prefer testing in CBR to avoid discussion/criticism on how the codec bitrates were calibrated for the test, but then again the codecs we generally test here are often used (and give slightly better quality) in VBR mode, so there's also some merit in testing them that way. You see, it's not that easy to decide.
I believe you should stay a bit longer here and read up on some more things before you accuse us of using unscientific methods.
Chris
Also, fix your biased testing methodology and keep the bitrates as close as possible to the original! Jesus. My hypothetical codec produces output bitrates of 192 kb/s when my quality setting intends it to be half as less ...
You have no clue what you're talking about, now do you?
Your unscientific results makes me keep clear of your website these days.
Dear Mr "8-posts", you're free to go wherever you want. Don't worry, we won't miss you either.
Have a nice day.
I'm guessing he's in violation of TOS #12, though I can't prove it at the moment.
Nevertheless, I must agree that these nonsensical and unfounded "contributions" are part of the problem rather than the solution. Our forum is better off without them. Please run along if you can't be constructive, Warning.
My hypothetical codec produces output bitrates of 192 kb/s when my quality setting intends it to be half as less, I demand it to be nominated as capable of transparency at 16 kb/s.
I failed to understand this sentence. My guess is like this:
In the last Opus 64k public test, the codecs and settings were calibrated to provide ~64kbps on a large variety of music, but these tests tend to use harder samples and Vorbis used 73.4kbps for the tested samples, which are typically hard. This means, Vorbis will use about 54.6kbps for easier samples, which is somewhat worrisome and annoys me but not something I'd consider unscientific.
However, if Vorbis allocate only 32kbps for easy samples people typically don't test, words like "abuse" or "cheat" would come up to my mind. This hypothetical hyper-VBR allocates more bits to the hard samples people typically test, and "abusing" easy samples exists in calibration albums but people don't test.
Ideally, average bitrates of test samples should be close to calibration albums by including enough easy samples, but preference for harder samples is unavoidable sometimes. My opinion is that testers calculate calibrate_albums_bitrate*2 - tested_samples_bitrate and make sure it isn't extremely low. When it is, add more easy samples.
Results:
ABC/HR Version 1.0, 6 May 2004
Testname: grcitrance MP4/OPUS
1R = M:\grcopus32.wav
2L = M:\grc64MP4.wav
3L = M:\grc64MP4cbr.wav
4R = M:\grc32MP4.wav
5L = M:\grcopus48.wav
6R = M:\grcopus64.wav
7R = M:\grc48MP4.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: M:\grcopus32.wav
1R Rating: 2.5
1R Comment: Extremely rough
---------------------------------------
2L File: M:\grc64MP4.wav
2L Rating: 3.6
2L Comment: Kinda duller and the upper freqs are distorted
---------------------------------------
3L File: M:\grc64MP4cbr.wav
3L Rating: 3.6
3L Comment: cant tell apart from sample2, I believe it's slightly better
---------------------------------------
4R File: M:\grc32MP4.wav
4R Rating: 2.4
4R Comment: Somewhat less rough than sample1 but has reduced stereo
---------------------------------------
5L File: M:\grcopus48.wav
5L Rating: 2.9
5L Comment: Less distortion than 1 but more than 2 and 3
---------------------------------------
6R File: M:\grcopus64.wav
6R Rating: 4.1
6R Comment: Noticeably rougher treble than original and slightly dull but otherwise pretty good
---------------------------------------
7R File: M:\grc48MP4.wav
7R Rating: 2.8
7R Comment: It has less distortion in the upper freqs than 5 but sounds duller like the upper freq was cut completely and stereo is reduced
---------------------------------------
ABX Results:
Original vs M:\grcopus32.wav
9 out of 10, pval = 0.011
Original vs M:\grc64MP4.wav
10 out of 10, pval < 0.001
Original vs M:\grc64MP4cbr.wav
10 out of 10, pval < 0.001
Original vs M:\grcopus64.wav
10 out of 11, pval = 0.006
M:\grc64MP4.wav vs M:\grc64MP4cbr.wav
3 out of 5, pval = 0.500
OGG sounds worse than Nero AAC, your own listening tests proved it before you started biasing the results.
I'm aware the tests are VBR, they STILL have to match the target bitrate as much as possible for the test to be scientific. All the bitrates in my test are VBR but they also are adjusted to be as close to 32, 48 or 64 kb/s as possible, I'm not gonna allow opus to go all the way up to 80 if the target is 64. Not too hard to comprehend. The truth is in the middle. Stop going to extremes and thinking CBR is your only other option. Do VBR and adjust the quality setting to match the bitrate. Stop thinking in black and white.
And I've been lurking here since 2007, thanks. I just rarely post. You run a tightly-moderated fascist hierarchy and are overrun with trolls and people with serious problems. I'll move to North Korea if I wanna deal with that.
I never use parametric stereo, I add -he to Nero. PS makes everything sound worse.
WMA superiority: http://en.wikipedia.org/wiki/Windows_Media...o#Sound_quality (http://en.wikipedia.org/wiki/Windows_Media_Audio#Sound_quality) I can't find the PDF at the moment, the link is outdated.
I'm guessing he's in violation of TOS #12, though I can't prove it at the moment.
And the fascistic impulse reveals itself already, didn't take too long I see. I wonder why many don't post here...
Achtung! Das ist sehr gut!
WMA superiority: http://en.wikipedia.org/wiki/Windows_Media...o#Sound_quality (http://en.wikipedia.org/wiki/Windows_Media_Audio#Sound_quality) I can't find the PDF at the moment, the link is outdated.
Thats WMA Pro, not WMA. They're different formats. WMA Pro is often compared here, but less often as it is essentially a dead format at this point.
Exactly. As poll 2013 has indicated that only two persons here use wma. If nobody uses it, nobody will test it. I'm not going to test it just for theory. It is a dead format.
Results:
Unless I'm misunderstanding something, you rated opus as the best quality for each bitrate?
Why do you think its worse then AAC?
I'm aware the tests are VBR, they STILL have to match the target bitrate as much as possible for the test to be scientific. All the bitrates in my test are VBR but they also are adjusted to be as close to 32, 48 or 64 kb/s as possible, I'm not gonna allow opus to go all the way up to 80 if the target is 64. Not too hard to comprehend. The truth is in the middle. Stop going to extremes and thinking CBR is your only other option. Do VBR and adjust the quality setting to match the bitrate. Stop thinking in black and white.
Thats not a very useful methodology IMO for testing VBR codecs. The entire point of VBR is to maximize quality for a given bitrate. Forcing a certain bitrate is fine, but at that point you might as well just use CBR as doing so defeats the purpose of VBR.
And I've been lurking here since 2007, thanks. I just rarely post. You run a tightly-moderated fascist hierarchy and are overrun with trolls and people with serious problems. I'll move to North Korea if I wanna deal with that.
FYI this kind of ranting makes you sound like you're insane.
OGG sounds worse than Nero AAC, your own listening tests proved it before you started biasing the results.
Show us this supposed “biasing”.
If anything, Vorbis has not been tested recently because of its having been overtaken by other formats (as was shown in 2011’s multi-format listening test (http://listening-tests.hydrogenaudio.org/igorc/results.html)) and its being in line to be replaced by Opus. Still, it is not particularly bad, as you seem to be trying to portray it. Perhaps you should redirect your little crusade against MP3 instead, since it does considerably worse and yet is significantly more popular. Then your frustrated energies might go somewhere.
And I've been lurking here since 2007, thanks. I just rarely post. You run a tightly-moderated fascist hierarchy and are overrun with trolls and people with serious problems. I'll move to North Korea if I wanna deal with that.
Yeah, you probably should stay away. At least then we’ll have one less of those trolls that you mentioned.
I also note your defaulting to the hoary old epithet
fascist, as if it’s anything other than a meaningless crutch for people with a chip on their shoulder who want to try to discredit someone who doesn’t agree with them with the maximal chance of getting approval from easily impressed onlookers.
Thankfully, though, the site does not go to the opposite extreme by being an enforced democracy; otherwise, ideas like yours in this thread would get much more traction than they deserve.
And the fascistic impulse reveals itself already, didn't take too long I see. I wonder why many don't post here...
OK! Bye now!
...and with this discussion the troll count just increased by one.
EDIT: I see the obvious irony was not overlooked in the previous post.
Please, how is accusing this forum of being biased and run by fascists be interpreted as anything other than trolling?
Unless I'm misunderstanding something, you rated opus as the best quality for each bitrate?
Only for 64 kb/s did it sound undeniably better, for the others I reluctantly rated it higher because I preferred the more dynamic stereo that opus retained at the expense of being more distorted. It was a choice of choosing from higher quality mono to lower quality stereo. I have another sample that suffers far harder from this distortion and sounds a lot worse than HE-AAC at the same bitrate and you don't even need to ABX to notice.
Why do you think its worse then AAC?
Because it sounds like garbage.
Thats not a very useful methodology IMO for testing VBR codecs. The entire point of VBR is to maximize quality for a given bitrate.
If your given bitrate is 64, don't let me see anything higher on the results page, or lower for that matter.
Forcing a certain bitrate is fine, but at that point you might as well just use CBR as doing so defeats the purpose of VBR.
Except CBR is a lot worse quality at the same output filesize than VBR. I guess the only reason people abandoned MP3 CBR is because they wanted to see what kinda strange numbers their encoder will pick in VBR? *facepalm*
FYI this kind of ranting makes you sound like you're insane.
Wasting your life on a website run by moderators with micropenis syndrome who run the place like their personal dollhouse sounds far more retarded to me. It's not surprising to see what kind of people it largely ends up with as exemplified by a post in the "what's your favorite format and why" poll by a guy who said he likes ATRAC the most because it feels warmest to him and then getting attacked by a pack of dorks for breaking some rule that forbids subjective opinions without an ABX test.......... in a thread where his subjective, meaningless opinion was outright asked for. Y'all need to chill out and take a pill sometime.
Wasting your life on a website run by moderators with micropenis syndrome who run the place like their personal dollhouse sounds far more retarded to me.
You’re really going all-out to be original, hilarious, and not at all thoughtless with your cosmopolitan array of epithets here, aren’t you? Boring gendered insults and accusations regarding mental health… It’s like I don’t even want to argue back because it just doesn’t seem worth it.
It's not surprising to see what kind of people it largely ends up with as exemplified by a post in the "what's your favorite format and why" poll by a guy who said he likes ATRAC the most because it feels warmest to him and then getting attacked by a pack of dorks for breaking some rule that forbids subjective opinions without an ABX test.......... in a thread where his subjective, meaningless opinion was outright asked for.
Nonsense. The question would have been in reference to objective features such as support in hardware/software, feature-set, licensing,
etc. To paint it as anything else is indicative of your having a seriously incorrect mental picture of and/or misrepresenting grudge against this site, which makes it your objections to it ever less tenable in either case.
Unless I'm misunderstanding something, you rated opus as the best quality for each bitrate?
Only for 64 kb/s did it sound undeniably better, for the others I reluctantly rated it higher because I preferred the more dynamic stereo that opus retained at the expense of being more distorted. It was a choice of choosing from higher quality mono to lower quality stereo.
If you don't think it sounds better, then you shouldn't have rated it higher.
IMO the problem here is that you either did the test wrong, or did the test right and don't agree with yourself. I think you need to figure out what it is you believe before posting such strong statements that may or may not be correct.
If your given bitrate is 64, don't let me see anything higher on the results page, or lower for that matter.
If you're testing CBR, sure. But for VBR that will not give you very meaningful results IMO.
Except CBR is a lot worse quality at the same output filesize than VBR. I guess the only reason people abandoned MP3 CBR is because they wanted to see what kinda strange numbers their encoder will pick in VBR? *facepalm*
Wait, so CBR is worse quality, but you think people shouldn't have abandoned it? Why would you recommend someone use settings that you believe don't work as well?
Wasting your life on a website run by moderators with micropenis syndrome who run the place like their personal dollhouse sounds far more retarded to me. It's not surprising to see what kind of people it largely ends up with as exemplified by a post in the "what's your favorite format and why" poll by a guy who said he likes ATRAC the most because it feels warmest to him and then getting attacked by a pack of dorks for breaking some rule that forbids subjective opinions without an ABX test.......... in a thread where his subjective, meaningless opinion was outright asked for. Y'all need to chill out and take a pill sometime.
I understand that you're angry, but silly rants like this don't help your cause. They just make you seem bigoted and ignorant. You need to calm down and then think carefully about what you want to say.
http://www.sendspace.com/file/x4ec7d (http://www.sendspace.com/file/x4ec7d)
AAC and opus at 32 kb/s VBR but both ended up at the exact same filesize. I'd include original wav but didn't feel like uploading 25MB so it's compressed at -q 0.75 and I think qualifies well enough as the original.
You don't even need to ABX to hear what complete shit opus sounds like with this sample. HE-AAC sounds almost transparent, though I'm sure an ABX would help me notice the flaws.
To the best of my knowledge, opus is not related to USAC which is emerging as the next future standard in audio to succeed AAC which I guess means opus will eventually be trashed by superior quality like OGG was by AAC and I hope it does.
EDIT: saratoga, re-read my post. Given your obvious trolling, you're on ignore.
.... I'd include original wav but didn't feel like uploading 25MB....
Why not compress it with FLAC and upload it?
You don't even need to ABX to hear what complete shit opus sounds like with this sample. HE-AAC sounds almost transparent
Aside from your usual exaggerated language, I agree with both of these points. So? It’s one sample. I’d like to think you’ve compared Opus with other codecs across a very broad range of samples and types of material to arrive at your current, authoritatively marketed conclusion.
And anyway, what is the ultimate purpose of this? ‘I don’t like Opus at these very low bitrates, therefore it should be abolished immediately; if you still don’t believe me, listen to this barrage of insults, and PLEASE PAY ATTENTION TO ME’
Whether you like it or not, many people want and will support a forward-looking open codec. Perhaps you should either ignore it or change your approach to aid its development constructively rather than continuing this apparent vendetta.
Or, if you don’t want to do that, I’m sure there are other sites where you can be consoled by others with a similar grudge; here, you’ll probably just get more frustrated, and something tells me you don’t need any more of that.
No one will refuse constructive feedback. But you have to approach it that way. You can’t just storm in and insult everyone and expect people to consider the finer aspects of what you’re saying. See how much delay that caused?
EDIT: saratoga, re-read my post. Given your obvious trolling, you're on ignore.
I think I'm raising fair points here. You can't just rate Opus higher quality in your listening test and decide that its lower quality. The fact that you have suggests that you have either done the test incorrectly, or that you in fact are confused about the relative quality of the samples. Surely you can see that contradictions like this indicate a problem?
.... I'd include original wav but didn't feel like uploading 25MB....
Why not compress it with FLAC and upload it?
Because I didn't feel like uploading 12.5MB.
I've examined the -q 0.75 MP4 with a spectrograph and it only has a couple holes in the very top 20khz shelf in the noise pouches between the spikes of the percussion, so even if you could hear 20khz it wouldn't have a loud enough intensity to be audible and even if you could still hear it, you're missing a few pouches of noise. Not a big deal.
I'll upload a FLAC if people insist, though.
AAC and opus at 32 kb/s VBR HE-AAC sounds almost transparent
Yikes, I'm glad I don't have those ears...
it's compressed at -q 0.75 and I think qualifies well enough as the original.
Really....?
AAC and opus at 32 kb/s VBR
I feel sorry for anyone that has to resort to 32kb/s for music...
I think I'm raising fair points here. You can't just rate Opus higher quality in your listening test and decide that its lower quality. The fact that you have suggests that you have either done the test incorrectly, or that you in fact are confused about the relative quality of the samples. Surely you can see that contradictions like this indicate a problem?
In the particular sample I tested, I didn't find opus lower or higher quality in the 32 or 48 bitrates, I just found it different, like the developers sacrificed sample quality to retain stereo dynamics. AAC could do that too. I rated opus samples higher reluctantly because they had better stereo in a song where sample quality wasn't as important as stable stereo.
The question is, how does Opus or any codec only have an advantage at one bitrate? It should be higher quality in all cases.
I think I'm raising fair points here. You can't just rate Opus higher quality in your listening test and decide that its lower quality. The fact that you have suggests that you have either done the test incorrectly, or that you in fact are confused about the relative quality of the samples. Surely you can see that contradictions like this indicate a problem?
In the particular sample I tested, I didn't find opus lower or higher quality in the 32 or 48 bitrates, I just found it different, like the developers sacrificed sample quality to retain stereo dynamics. AAC could do that too. I rated opus samples higher reluctantly because they had better stereo in a song where sample quality wasn't as important as stable stereo.
According to your test results, you did find it higher quality at all bitrates, hence its a little strange that you're trying to claim otherwise. If you really think that the Opus sample did worse, then you screwed up the test.
The question is, how does Opus or any codec only have an advantage at one bitrate? It should be higher quality in all cases.
There is no reason to think this. Different codecs work better at different bitrates. For example, at 32kbps I'd strongly prefer AAC-HE to MP3. At 192kbps, I prefer MP3.
Furthermore, for any given sample, two equally good codecs will tend to do better or worse then each other based on chance. Usually if you want to find out how two codecs compare, you pick a number of samples and evaluate all of them and then average the scores.
Whether or not I’m being automatically ignored here, hooray, I’m genuinely glad that actual audio technology is being discussed in a way that might well be constructive, and I have a reference to cite in the future as an illustration of why that works and posting whilst irate does not.
I've examined the -q 0.75 MP4 with a spectrograph and it only has a couple holes in the very top 20khz shelf in the noise pouches between the spikes of the percussion, so even if you could hear 20khz it wouldn't have a loud enough intensity to be audible and even if you could still hear it, you're missing a few pouches of noise.
I guess we should start encoding with Blade @128kbits for archival quality.
The question is, how does Opus or any codec only have an advantage at one bitrate? It should be higher quality in all cases.
Remember, Opus is a very-low-delay format. HE-AAC or xHE-AAC (USAC) are not. So why are you surprised that Opus sounds worse on this very tonal sample? It shouldn't be surprising that for music, low-delay formats lose efficiency/quality much quicker at low bitrates than higher-delay formats. By the way, have you compared Opus and (HE-)AAC at 96 or 128 kbps?
Really....?
I guess we should start encoding with Blade @128kbits for archival quality.
Come on, guys, that -q 0.75 (292 kbps average) certainly serves well enough to get an idea of the original recording. Do we have to be
that picky? The test encodings are less than 32kbps stereo!
Chris
Remember, Opus is a very-low-delay format. HE-AAC or xHE-AAC (USAC) are not. So why are you surprised that Opus sounds worse on this very tonal sample? It shouldn't be surprising that for music, low-delay formats lose efficiency/quality much quicker at low bitrates than higher-delay formats. By the way, have you compared Opus and (HE-)AAC at 96 or 128 kbps?
By default Opus uses 20 millisecond frames. I wonder if increasing it to 40 ms helps for tonal samples like this, or if this is just one of those samples where SBR is too well suited.
@Chris:
That's not the point. We don't judge quality from spectral plots.
While I'm happy the OP has shown enough respect to supply some DBT results (even though they don't support the claims), this makes me wonder whether this discussion will ever go anywhere. Maybe my leaving it will help.
@seren
I don't use 32 kb/s for music, but I'm hoping it will happen someday. The original format of the song I uploaded is .MOD and it's only 15KB which gives it a bitrate of 1.2 kb/s and even lower when compressed.
@saratoga
I've explained myself numerous times, I won't do it a third. Also, choosing MP3 for quality isn't a wise choice given its inefficient block sizes and other numerous design flaws make it incapable of transparency at any bitrate.
@db1989
You're just being ignored, for now anyway. I intended to respond but posts concerning the original topic came back up so I went with the flow.
@greynol
I never said anything about archival quality.
@C.R.Helmrich
I increased Opus' delay to the maximum 60ms to produce that sample. Is this close to AAC's? I'm aware it was intended for VoiP but your latest listening test got my attention when I heard AAC has finally been made obsolete, so I was puzzled why this was only true for one bitrate? No I haven't ABX'd higher bitrates but since I do not intend to switch to Opus even if it is only slightly higher quality, I'm not wasting the time. USAC is due which will crush both AAC and Opus.
@saratoga
I've explained myself numerous times, I won't do it a third.
I don't think two counts as "numerous", particularly when one of them is you accusing me of trolling. Basically, you've just said that you rated Opus higher even though you don't think you should have. Is that a fair summary?
Also, choosing MP3 for quality isn't a wise choice given its inefficient block sizes and other numerous design flaws make it incapable of transparency at any bitrate.
No, MP3 is quite transparent at high bitrates. You need to do some listening tests, I suspect you'll be surprised. Or alternatively, take a look at some of the historical listening tests on this site.
I increased Opus' delay to the maximum 60ms to produce that sample.
Hmm, did you use any other non-standard settings? Where is that FLAC file you mentioned, I'm curious to try it myself.
I'm aware it was intended for VoiP but your latest listening test got my attention when I heard AAC has finally been made obsolete, so I was puzzled why this was only true for one bitrate?
Did you see my explanation above? You shouldn't assume that one codec will be better at all bitrates. AAC-He is a great example of a codec that does really good at low bitrates and not good at higher bitrates.
I don't think two counts as "numerous", particularly when one of them is you accusing me of trolling. Basically, you've just said that you rated Opus higher even though you don't think you should have. Is that a fair summary?
No, opus had better stereo dynamics at the expense of sample quality. AAC had better sample quality with the expense of simplified stereo. They both sounded like shit in the end and which one's worse is a matter of opinion. Do you prefer better brightness or better stereo?
No, MP3 is quite transparent at high bitrates. You need to do some listening tests, I suspect you'll be surprised. Or alternatively, take a look at some of the historical listening tests on this site.
Already have. Trained ears can distinguish the MP3 artifacts at even 320 kb/s due to serious design flaws, even if ours can't. Both will sound transparent at 192 kb/s to both of us and AAC will have better technical quality, retaining more inaudible, high-frequency parts so what's the point of using an outdated format if you're aiming for quality, pal?
Hmm, did you use any other non-standard settings? Where is that FLAC file you mentioned, I'm curious to try it myself.
Nope. I'll upload a FLAC later if anyone else requests.
Did you see my explanation above? You shouldn't assume that one codec will be better at all bitrates. AAC-He is a great example of a codec that does really good at low bitrates and not good at higher bitrates.
Except it does just fine and outperforms MP3 at any bitrate. For real dude, you aren't too bright for someone that's been here for a decade. I've learned everything I needed about AAC within a month when I decided to switch to it. This site's listening tests also prompted my move when it kept being rated #1 even if y'all now downsize its bitrate, fraudulently advertise it as higher and declare it to be lower quality than f*cking QUICKTIME.
I don't think two counts as "numerous", particularly when one of them is you accusing me of trolling. Basically, you've just said that you rated Opus higher even though you don't think you should have. Is that a fair summary?
No, opus had better stereo dynamics at the expense of sample quality. AAC had better sample quality with the expense of simplified stereo. They both sounded like shit in the end and which one's worse is a matter of opinion. Do you prefer better brightness or better stereo?
That is not what you said before:
So I compared at different bitrates and opus always sounded worse.
And then you rated Opus as having higher quality. So you've basically said that Opus is better, that both are equally bad, and that AAC is better.
Like I said before:
think you need to figure out what it is you believe before posting such strong statements that may or may not be correct.
And I still mean it.
Already have. Trained ears can distinguish the MP3 artifacts at even 320 kb/s due to serious design flaws, even if ours can't.
Source? This sounds like nonsense to me.
Both will sound transparent at 192 kb/s to both of us and AAC will have better technical quality, retaining more inaudible, high-frequency parts so what's the point of using an outdated format if you're aiming for quality, pal?
What is "technical quality"? How does one assess it?
Nope. I'll upload a FLAC later if anyone else requests.
Requested. Upload it now so I can try.
Except it does just fine and outperforms MP3 at any bitrate.
This I doubt. AAC-He is not really intended to be transparent. Its basically just estimating higher frequencies via harmonic extension and some side information. But by all means, if you have evidence, I'll look at it now.
For real dude, you aren't too bright for someone that's been here for a decade. I've learned everything I needed about AAC within a month when I decided to switch to it.
I'm not attacking you, so no need to get so worked up over nothing.
This site's listening tests also prompted my move when it kept being rated #1 even if y'all now downsize its bitrate, fraudulently advertise it as higher and declare it to be lower quality than f*cking QUICKTIME.
Not sure if you're aware of this and just mistyping, but Quicktime is an AAC encoder, and in fact it it wildly considered to be among the best AAC encoders.
AAC will have better technical quality, retaining more inaudible, high-frequency parts
Hey Chris,
Are you OK with this? I'm not!
That is not what you said before:
So I compared at different bitrates and opus always sounded worse.
And then you rated Opus as having higher quality. So you've basically said that Opus is better, that both are equally bad, and that AAC is better.
I said I COMPARED at different bitrates, not ABX'd. That sample was the only one I ABX'd because I couldn't tell immediately by listening if Opus had an advantage. Stereo defined that particular song more than brightness. I already explained this before.
Like I said before:
think you need to figure out what it is you believe before posting such strong statements that may or may not be correct.
And I still mean it.
I don't wanna believe. I'm not interested in being right, I'm interested in being correct. A concept OGG-pimping cock-smokers especially fail to grasp or any other codec fanboys for that matter.
Source? This sounds like nonsense to me.
I'll let you know when I find it. I recall some people on here boasting of such abilities too.
What is "technical quality"? How does one assess it?
Behold the awesome spectrograph, showing human eyes what human ears can't:
original (http://i48.tinypic.com/261yntj.png)
MP3 192 (http://i46.tinypic.com/343n7ki.png)
MP4 192 (http://i46.tinypic.com/2zfpc8z.png)
Requested. Upload it now so I can try.
I said if someone ELSE also requests it. Here you go anyway, even if you don't deserve my clemency, blatant troll. (http://www.sendspace.com/file/2b5pyt)
This I doubt. AAC-He is not really intended to be transparent. Its basically just estimating higher frequencies via harmonic extension and some side information. But by all means, if you have evidence, I'll look at it now.
You're confusing an extension with an encoder. MP3 can use SBR too, it's called mp3pro. It still sucks and sounds worse than AAC.
I'm not attacking you, so no need to get so worked up over nothing.
You said CBR and VBR MP3 at the same filesize is the same quality. Something only a colossal mouth-breather would say after being exposed to audio tech resources for a decade, or a troll playing dumb like you.
Not sure if you're aware of this and just mistyping, but Quicktime is an AAC encoder, and in fact it it wildly considered to be among the best AAC encoders.
So why was it surpassed every single time by Nero in the past and now when you guys decide to use a considerably lower bitrate for Nero like here (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/results.html) it ends up last. Well if that just wasn't predictable... Biased testing to the core.
TOS 8 and 2.
Your posting style (it is painfully obvious who is the troll here) and insistence that frequency response matters is unwelcome. Tone it down or you're gone.
Regarding your link, Apple tvbr had a lower bitrate than nero.
I was tempted to reply to this thread yesterday but I opted to wait and calm down. It is obvious that the thread hasn't really advanced.
Warning is clearly biased and has been confirming it in each of his posts.
- First biasing shown: Opus, being free, open source, and being endorsed by the same team than Vorbis means it's a second grade codec.
- Second biasing shown: AAC, being an MPEG codec, is a first grade codec.
- Addenum to second biasing: AAC is an evolution of MP3 to improve it, HE-AAC is an evolution of AAC to improve it, USAC will be an evolution of AAC improve it.
- Third biasing shown: A better codec has a higher frequency response spectogram (we are at Hydrogenaudio, so we won't accept that as a proof no matter the insistence in doing so).
- Addenum to third biasing: high frequency response is synonym to quality (This is the reason why judging HE-AAC and Opus at 32kbps, the OP says Opus is rough, worse, sounds like garbage... Even when the only sample he tested and provided an ABX, he valuated the stereo separation of Opus making it outperform HE-AAC)
- Possible fourth biasing: Golden ears exists, and I believe them. (As shown by the sentence Trained ears can distinguish the MP3 artifacts at even 320 kb/s due to serious design flaws, even if ours can't.)
All that, without counting the several times he has taken the opportunity to show disparagement, and other bad behaviours, like in post #8 (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=99084&view=findpost&p=822072).
So my question is this: Don't we have enough of this already?
It didn’t take long for the nose-thumbing tantrum to resume. I’ll let someone else decide, but unless Warning stops being so exaggeratedly incivil, I suggest shutting the whole thing down.
Anyway, do you think your little rants are impressing anyone here? I really doubt it. At worst, you might be able to link similarly biased trolls here and have a good chuckle with them, but as long as they don’t start posting like you, I couldn’t care less.
Actually, having read the latest addition to the Recycle Bin since I wrote the above, I don’t see the point of continuing to provide a platform for this hateful attention-seeker. At worst, he gets to cry INTERNET FASCISM to other people who have his same ridiculous approach to forums. So what? A couple more angry and biased trolls in the world. Boo hoo. By this point, you’re a drop in the ocean. Your contributions here are doing nothing but prove why we need rules like those we already have.
But I’m going to leave this open for now to see whether anyone else wants to have a punt. I wouldn’t want to oppress their free speech, after all!
Good morning,
I increased Opus' delay to the maximum 60ms to produce that sample. Is this close to AAC's?
This I doubt. AAC-He is not really intended to be transparent.
OK, some clarification is necessary here.
HE-AAC has a frame length of 2048 samples in the normal dual-rate mode, and with its 50% window overlap, one frame spans 4096 input samples = 93 ms @ 44.1 kHz. Opus has very little frame overlap, which makes efficiency go down on tonal signals. So even 60-ms windows in Opus wouldn't be enough to get to the efficiency Level of HE-AAC. And IIRC Opus doesn't have 60-ms
windows. I think the developers once mentioned here that the CELT
block size can never be more than 20 ms, so a 60-ms frame just contains 3 blocks. Yes, if the block size were 40 ms, it would probably sound better.
HE-AAC is in fact intended to be transparent. In downsampled SBR mode, you can move your SBR start frequency up to 16 kHz or more, i.e. into (nearly) inaudible regions. But nobody uses it that way because then it gives no advantage over AAC-LC.
AAC will have better technical quality, retaining more inaudible, high-frequency parts
Hey Chris,
Are you OK with this? I'm not!
[/quote]
If you define technical quality as e.g. "maintaining efficiency on different input signal characteristics", I'm perfectly OK with this. The rest is mostly a sign of encoder tuning and has nothing to do with the particular formats.
Some last words to Warning: please stop the cursing! What's the point of calling people trolls etc.
And please upload a FLAC of 30 seconds of that MIDI-like item. And upload it in the Upload (http://www.hydrogenaudio.org/forums/index.php?showforum=35) forum.
Chris
However, if Vorbis allocate only 32kbps for easy samples people typically don't test, words like "abuse" or "cheat" would come up to my mind. This hypothetical hyper-VBR allocates more bits to the hard samples people typically test, and "abusing" easy samples exists in calibration albums but people don't test.
Ideally, average bitrates of test samples should be close to calibration albums by including enough easy samples, but preference for harder samples is unavoidable sometimes.
Including easy samples in the test should be desirable since the codec should be equally challenged in making a level of transparency with less bits; really testing how effective the VBR is.
So why was it surpassed every single time by Nero in the past and now when you guys decide to use a considerably lower bitrate for Nero like here (http://listening-tests.hydrogenaudio.org/igorc/aac-96-a/results.html) it ends up last. Well if that just wasn't predictable... Biased testing to the core.
What?!
Nero 95-96 kbps
Apple TVBR 93-94 kbps.
And still Apple AAC encoder was much better than Nero. It was actually day and night difference.
Instead of posting harsh and uneducated comments You could read rules of this forum.You're confusing us with some "audiophool" community.
USAC is due which will crush both AAC and Opus.
How naive, how naive.
MPEG Surround was here since 2007. And there is still no available codec in 2013.
It was supposed that it would be superior to HE-AAC/HE-AACv2 at <32 kbps (music) in that moment.
Now,2013, who is interested in <32 kbps (music)?
So while waiting for a USAC real encoder take a deep breath and count until billion (and a little bit too)
USAC has +8 kbps advantage over AAC, and beleive it or not, it means that USAC and Opus will have a same quality for 64 kbps and higher. What about <64 kbps? It's not 2003 when people were interested in 32 kbps audio. It's 2013 and 64 kbps is considered comfortably low. Poll (http://www.hydrogenaudio.org/forums/index.php?showtopic=72265).
Even YouTube uses 96 kbps audio for default quality video streams, and =>128 kbps audio for HD.
Instead of spitting googled information about USAC, here some people already have some real stuff and know what will be happening in a few next years.
I wouldn't mind about a permanent ban.
I'm aware the tests are VBR, they STILL have to match the target bitrate as much as possible for the test to be scientific. All the bitrates in my test are VBR but they also are adjusted to be as close to 32, 48 or 64 kb/s as possible, I'm not gonna allow opus to go all the way up to 80 if the target is 64. Not too hard to comprehend. The truth is in the middle. Stop going to extremes and thinking CBR is your only other option. Do VBR and adjust the quality setting to match the bitrate. Stop thinking in black and white.
It sounds like what you want is ABR ("average bit rate", every file as close as possible to nominal) rather than VBR (close to nominal rate averaged over a wide variety of material). WIth ABR, relative ranking of codecs could change if you test an album coded as one file vs individual tracks.
For my money, vbr would give the best overall bang for the bit on my player, and also (afaik) encode faster since the encoder doesn't have to analyze the whole song before it gets down to business (or in your case, encode, adjust, repeat until you get precisely 64k).
To summarize, smarter VBR is an advantage in real performance. If you cripple that, the test is less valid.
If you define technical quality as e.g. "maintaining efficiency on different input signal characteristics", I'm perfectly OK with this. The rest is mostly a sign of encoder tuning and has nothing to do with the particular formats.
Ok great, Blade has a very high degree of technical quality.
Ok great, Blade has a very high degree of technical quality.
AFAIR, Blade was inefficient on almost every kind of input signal characteristic Btw, I was referring to the format (he said "AAC will have better..."), not some implementation of it.
Chris
You better go back and re-read. My Blade comment was (and still is) fair game.
This really boils down to nothing more than passing-off lossy as lossless because the spectral plot is nicely filled-in:
I've examined the -q 0.75 MP4 with a spectrograph and it only has a couple holes in the very top 20khz shelf in the noise pouches between the spikes of the percussion, so even if you could hear 20khz it wouldn't have a loud enough intensity to be audible and even if you could still hear it, you're missing a few pouches of noise.
What a wonderful display of ignorance!
My encoded AAC is much worse than yours. And that is very strange...
I'm posting to announce I've done ABX listening tests to... confirm what I already knew and hence all your trolling has been successful.
On a tough-timbred song I tested different MP3 bitrates at mono to not deal with the stereo-spacing bullshit like with opus vs. Nero, up to 192 kb/s. On 128 kb/s I got 37/50, 28/40 for 160 and 37/60 for 192, and I do not even have golden ears.
Since this is a mono file, 192 is more or less equivalent to 384 for a stereo song, well above the 320 kb/s max limit. MP3 blows, it was never meant to be transparent. Deal with it and move on.
Nero AAC on the other hand was transparent at 80 kb/s.
I took the liberty of also testing QTAAC to see if you guys really aren't full of shit with your biased listening tests. Its horrible flexibility made it an unmanageable bitch to do a fair comparison. On the same mono file it failed to keep the 44.1 samplerate on the lower bitrate I selected and failed to produce CBR at CBR mode and ended up with 44 kb/s when 40 was selected. Nero had none of these issues.
So naturally QT at 32 khz with the samebitrate vs Nero at 44 khz, it beat Nero.
Encoding with Nero at samplerates below 44 is not an option either as Nero for some reason sucks capital ass with these samplerates. To avoid a very frequent misinterpretation with you illiterate fucks: Nero produces much lower quality at lower samplerates with the SAME bitrate. Yes, tell the devs to fix that shit. MP3 even can compete with AAC if it uses a 32khz samplerate at 128 kb/s and majority of post-pubescent people can't hear above 32khz anyway.
Anyway, testing with a real, stereo song at 64 kb/s with Nero and QT, QT was audibly worse than Nero. Yeah, Apple's audio gear sucks as much dick as their H.264 implementation that ended up being worse quality than MPEG-fucking-2. Embarassing company.
About Blade MP3, I downloaded it to test greynol's bullshit assertion. Surprise, surprise. Not only does it NOT retain higher frequencies than AAC, the degradation of the formants was so obvious on the spectrograph that I could predict immediately the horrible flanging that would result if I hit the play button... and out the defects came. As obvious on the ears as it was on the eyes.
To actually think that I expected any kind of intelligence from this troll site is laughable, I must be insane.
But as I'm an honorable guy, I will hand over ABX results and spectrogrpah screenshots (Blade) upon request and that's the last I'll have of your nazistic dump.
Or you can silence this post too, as we all know your needle-dicked members of authority hate being contradicted.
Btw, did another opus vs. AAC test with that 'tonal' sample as you call it. opus didn't disappoint me:
ABC/HR Version 1.0, 6 May 2004
Testname: Mazgal AAC/opus
1L = M:\mazgal64MP4.wav
2R = M:\mazgal32opus.wav
3R = M:\mazgal64opus.wav
4L = M:\mazgal64MP3.wav
5R = M:\mazgal32MP4.wav
6L = M:\mazgal48MP4.wav
7L = M:\mazgal48opus.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: M:\mazgal64MP4.wav
1R Rating: 4.8
1R Comment: The strike of the heavy notes have a little treble accentuation, difficult to notice
---------------------------------------
2R File: M:\mazgal32opus.wav
2R Rating: 2.4
2R Comment: Extremely metallic melody
---------------------------------------
3R File: M:\mazgal64opus.wav
3R Rating: 4.0
3R Comment: A much reduced form of distortion than 2
---------------------------------------
4L File: M:\mazgal64MP3.wav
4L Rating: 2.2
4L Comment: Heavy smearing, top shelf gone, obviously 64kbps MP3
---------------------------------------
5R File: M:\mazgal32MP4.wav
5R Rating: 3.3
5R Comment: Stereo greatly simplified, some distortion on upper shelf and kinda echoey melody
---------------------------------------
6L File: M:\mazgal48MP4.wav
6L Rating: 3.6
6L Comment: Greater distortion in the treble than 5 but much less loss of stereo
---------------------------------------
7L File: M:\mazgal48opus.wav
7L Rating: 3.5
7L Comment: Seems to be more distorted than 6 but with more texture. Not sure what to decide.
---------------------------------------
ABX Results:
Original vs M:\mazgal64MP4.wav
18 out of 20, pval < 0.001
Original vs M:\mazgal64opus.wav
10 out of 10, pval < 0.001
Original vs M:\mazgal32MP4.wav
10 out of 10, pval < 0.001
Original vs M:\mazgal48MP4.wav
10 out of 10, pval < 0.001
Sup Warning. Bye Warning.
P.S. I hope the terrible attitude made abundantly clear in your post, like that of a foul-mouthed child who didn’t get enough lollipops, speaks for itself without me having to point out what a pitiable person you are. Oh – oops.
Actually, you know what? I’m moving this back out of the Recycle Bin to let people discuss the claims made in your post if they want to and can be bothered to navigate around all the stupid insults you littered it with. I won’t, however, change my mind about banning your pathetic ass (both cheeks, er, I mean accounts) for this laughable [this was wrong: it’s really not funny] tantrum of yours. You can cry censorship all night long if you want. We’re happy to censor people who conduct themselves as you have.
And nothing of value was lost.
This thread is generally awesome.
I mostly _expect_ that for music (speech is another matter) at very low rates where everything sounds like crud that Opus will be a little more cruddy than HE-AAC as just one of the inevitable consequences of our decision to have latency suitable for interactive use— an order of magnitude lower than HE-AAC. Certainly for some samples it will be the case. But low latency means wider adoption (by people who want the low latency) and lower decoder memory usage that will benefit people who don't care about latency. I hope and expect future Opus encoders to do somewhat better at those low rates (in part by taking advantage of additional lookahead when the delay is acceptable)... and that it might be helpful if people collected examples where Opus was unusually bad so we'd have them for testing when we got around to paying attention to low-rate music. (Rearranging the deck chairs on the inadequate rate titanic hasn't been the greatest priority)... and in any case, people targeting low rate music who don't care about Opus' other qualities (latency, speed of the OSS encoder, licensing, scalability, speech performance)... might prefer to use other codecs. And that's okay. Choice is good.
But no, the results posted— as noted by Saratoga— actually give Opus a higher score at each rate tested, including the lowest one.
We live in a mad mad world.
(Edit: I actually missed the second page of this thread until now. Weee... Well using large frame sizes in Opus will not help: They will only save you on the order of a few hundred bit/s except in the LPC mode— and people aren't talking about rates low enough where the LPC mode would be used on music)
About Blade MP3, I downloaded it to test greynol's bullshit assertion. Surprise, surprise. Not only does it NOT retain higher frequencies than AAC, the degradation of the formants was so obvious on the spectrograph that I could predict immediately the horrible flanging that would result if I hit the play button... and out the defects came. As obvious on the ears as it was on the eyes.
I see my point was completely missed. I guess this happens when all you want to do is be an ip-shifting internet troll.
For those who haven't seen it and would like to be amused (please don't talk about it here though, m'kay?!?):
http://www.head-fi.org/t/225356/lossy-audi...s-update-on-p-7 (http://www.head-fi.org/t/225356/lossy-audio-codecs-comparison-huge-amount-of-pics-itunes-update-on-p-7)
How can they compare audio files by images which were compressed with jpeg, I don't understand. This is a transcode.
Part of me would like to enumerate all the ways in which I’m bothered by that thread and the absurd abuses of logic committed by the OP and his co-thinkers, but I don’t think it’s a good idea to put myself or anyone else through that. Suffice it to say that it makes me even more keen to find a way off this planet. I agree with greynol that we shouldn’t discuss it, both because it’s not worth it and because we’ve had enough OT and troll-baiting around here in the last week or so without heaping more on top.
Woo. I could feel the hairs growing on the palms of my hands as I read the thread. It did make me wonder about something:
I understand that one of Opus' design aims is that it is a very low latency format, optimised for interactive applications such as VOIP. This implies optimisation for applications involving voice (speech). Does anyone know of a low latency interactive application that would require optimisation for other content such as music?
I realise it's not specifically an Opus question, so if it's likely to involve more than a short answer feel free to ask me to start another thread. I just thought I'd ask while the people most likely to know the answer were "in the room"...
Nero produces much lower quality at lower samplerates with the SAME bitrate. Yes, tell the devs to fix that ****
Nobody works on Nero AAC encoder anymore since quite a long time ago
Yeah, Apple's audio gear sucks as much dick as their H.264 implementation that ended up being worse quality than MPEG-*******-2 ...
It depends of bitrate. A high quality MPEG-2 encoders like HC, Carbon Coder and particularly CCE SP3 will do an excelent job at high rates (Blu Ray, 720-1080p 20 Mbits and more). Of course, Apple H.264 encoder isn't near as good as x264 but none of those MPEG-2's would compete with it, let's say, at 3 Mbps 720p.
I'm not one of Apple followers (and actually more into Samsung devices ) but, clearly, they have damn good AAC coder. As of Nero, I wish it was any good as other AAC encoders because it was (maybe still is for some people) actually a "free AAC coder for folks". As of HE-AAC(v1/v2) encoders, imo Fraunhofer is the best one.
Or you can silence this post too...
Everything is in contrary.
It's all about speaking up loudly, freely and civilized manner.
I understand that one of Opus' design aims is that it is a very low latency format, optimised for interactive applications such as VOIP. This implies optimisation for applications involving voice (speech). Does anyone know of a low latency interactive application that would require optimisation for other content such as music?
Internet Jam Sessions (or more broadly Networked Music Performance (http://en.wikipedia.org/wiki/Networked_music_performance)) plus speech chat with background music and/or simultaneous music sharing or tuition and other non-speech audio including remote fault diagnosis where transparent non-speech sound can be helpful. Also wireless headphone links (esp when synchronised with images, demanding low latency). For more suggestions, see Wikipedia's Opus (audio format) (http://en.wikipedia.org/wiki/Opus_%28audio_format%29) article, which now includes numerous citations for such applications.
At CELT rates rather than SILK, Opus is also less prone to false voice effects caused by background noise - e.g. sometimes when Skyping to collaborate on some work, while not actively chatting it will pick up creaks and clunks and shuffling paper and make them sound like voices (reminds me of EVP machines (http://en.wikipedia.org/wiki/Electronic_voice_phenomenon)!)
Thank you for the explanation. I've noticed similar sound effects when using the VOIP audioconferencing systems at the office.