xiph
Pretty significant artifacting even at 320kbps and 384kbps
So...
Opus doesn't seem to like a lot of electronic music... on this sample in particular it's having issues...
Here's the lossless
https://drive.google.com/file/d/1ERvzuTO86OcSTBcI7uOLlrY5dGaTE8Px/view?usp=sharing
Especially near the beginning and middle... near the end the music changes and I don't think it's as audible (but then again, I didn't test the end of the sample much... as I was hearing a lot of artifacting in the beginning and middle)
At 48kbps it has pretty much the same artifacting but MUCH more audible... it's mainly in the 0-1khz range... same with 320kbps
This is the sample encoded with Opus at 48kbps
https://drive.google.com/file/d/1DQrGaXP82Wx1H20nlPKtGY7Fts91i6nh/view?usp=sharing
It's whirling/thumping sounds that I'm hearing... mostly artifacts in the 0-1khz range... at 48kbps it's understandable... but at 320.... not so much
Oh, and here's the ABX test of 320kbps Opus versus lossless
https://i.imgur.com/FIsTWC4.png
It seems to be issues with mostly electronic music... at least, I think...
I used 1.2.1 Opus and VBR
As a comparison to AAC at 48kbps... you can tell it is a lot better because it doesn't have those whirling effects... or at least, drastically reduced
https://drive.google.com/file/d/1R36tsP0NrSxFvekfyeWXSzFMTK_YzkbV/view?usp=sharing
Now, AAC did use about 7% more bitrate overall... but even then... it's a LOT better
I wanted to be sure it wasn't just because I was listening to a short sample... I'm pretty sure the first 45 seconds or so there can be more artifacting.. haven't fully tested that... Anyway, I believe it was a lot harder (I can't remember how it was before) it took a lot longer and I only got 15/16.. I know I heard artifacts it was just.. EXTREMELY subtle.... Still, I think 256 or AT LEAST 288 or AAC is FULLY transparent.. so... 320 of Opus...
ABX test
https://i.imgur.com/Ry4HIIX.png
Here's with a different song.. this one was easy to ABX even at 384... I didn't ABX 448 but I did listen to it and a lot of the artifacting was gone... the thump from the 384 is completely gone and a lot of the crackling is gone... those were the artifacts... I briefly tried WASAPI and still heard the artifacts.... so it's not a DirectSound thing
https://i.imgur.com/XOXTzft.png
And the encoder settings
General Complete name : C:\Users\Bryce\Downloads\Aurelleah - Chromatica-test384.opus Format : Ogg File size : 11.9 MiB Duration : 4 min 20 s Overall bit rate : 385 kb/s Track name : Chromatica Performer : Aurelleah Recorded date : 2016 Writing application : opusenc from opus-tools 0.1.10 ISRC : GBSMU2645492 Comment : Visit http://aurelleah.bandcamp.com ENCODER_OPTIONS : --quiet --bitrate 384 --vbr --ignorelength
Audio ID : 27744 (0x6C60) Format : Opus Duration : 4 min 20 s Channel(s) : 2 channels Channel positions : Front: L R Sampling rate : 44.1 kHz Compression mode : Lossy Writing library : libopus 1.2.1
I made sure to normalize the lossless down -5db first and then recompress into Opus to avoid any further clipping, which would void the test
Here's the song for the last one... the 320 mp3 is available for free so I'm sure for testing purposes releasing the lossless is okay
It happens pretty much the whole song but there's some areas that are worse than others... like between around 1:45-1:55 in the song is one of the spots
This is the normalized version I recompressed with Opus... so that's why I'm linking it
https://drive.google.com/file/d/1ApIZ8h31R_Rtxr5NLJGk4GU6DVG4b3lD/view?usp=sharing
Your normalized flac seems to be an upsampled MP3: Here is the log from Audiochecker v2.0:
AUDIOCHECKER v2.0 beta (build 457) - by Dester - [email protected]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Started at: Montag, 30. 04. 2018. - 19:12.15
1 file found
1 -===- C:\Users\damia\Downloads\Chromatica [Melodic Electrostep]-normalized.flac
Extracted successfully
Conclusion: this track is MPEG with probability 95%
Tempfile successfully deleted.
Finished at: Montag, 30. 04. 2018. - 19:12.34 (operation time: 0:00.19)
Like you I came to the conclusion that HE-AAC (using Apples CoreAudio encoder) sounds better than Opus (using Opus Audio Encoder v1.2.1) at low bitrates in most scenarios (tested it with overall target bitrates (variable) of 64 kbps and 80kbps using complex music (metal and classical music)).
Hmm... weird
Maybe the software he used to create it were a bunch of mp3s shrugs
What about the first audio sample?
Because even 320kbps shouldn't have pretty drastic artifacts....
I could ask him what software he uses... but on his Youtube page it says
DAW: FL Studio 12 Synths: Serum, Sylenth1 Orchestral: EWQL Hollywood Orchestra, Metropolis Ark Piano: Pianoteq 5 D4/K2
artifactingthing-lossless.wav seems to be completely lossless.
Could you upload these 320kbps/384kbps Opus samples with "pretty drastic artifacts"?
Okay... I'll upload the 320kbps one since the artifacting is like 3x more audible versus 384... I can hear it at 384 but it is relatively subtle
320kbps with the Chromatica one: https://drive.google.com/file/d/1j9G2yMXN7ZDWwpW-Odmqz2szwH9xUZwX/view?usp=sharing
I find it especially noticeable between about 1:45 and 1:50 in the sample, at least.... but it happens throughout the song... it's basically thumping/popping sounds.. there was one part where there's a loud 'thump' but I can't remember where it is.. I could listen through it and find out
320kbps with the Giggle one: https://drive.google.com/file/d/1NvOUhT40GJx8mluNetTXwFx8O4-XuguT/view?usp=sharing
and lossless so you can compare: https://drive.google.com/file/d/1YgKhrWXyhWLgv-VFD3_9qwT6ENAm_dC7/view?usp=sharing
Now that one.. is a lot harder to detect the artifacts... if you can't hear them I'll link the 192kbps Opus one as it's M U C H more audible... but I think it's most audible around 1:00 to 1:15 or so.. I would link a short sample but it seems if enough time doesn't go on within an audio clip or something there will be a lot more artifacting.. it's weird... so that's why I'm linking the whole song
The artifacts with the first one is basically thumping/popping sounds..... the second one is mostly a whirling sound that makes the audio a lot less 'clean' sounding
On the first one there was a big thump that made it V E R Y easy to pass the ABX test even at 384kbps.... I should probably listen to it and try to find it again...
I can't find the thump.. but the crackling mostly happens between about 1:45 and 1:48.... it happens other places too, but that's one part where it's REALLY audible even at 384kbps
I should probably post what I'm using to listen...
I'm using: https://www.oppodigital.com/headphones-pm-3/
With: https://www.oppodigital.com/headphone-amplifier-ha-2SE/
except I'm using the older HA-2 version and not the HA-2SE
I'm not using 'bass boost' or anything... just the flat frequency response...
I realize I probably don't need an amp but... eh..
So.. I have a question, how does it detect whether something is truly lossless or not? I mean... I know it's not an absolute 100% sure that the song isn't truly lossless but still..
I don't know how Audiochecker (got it from this site: https://archive.org/details/Audiochecker.v2.0.Beta.Build.457) works in detail, but I found this chromatica piece in fully lossless FLAC here: https://pony.fm/tracks/21282-chromatica-future-dash-theme-melodic-electrostep. And here Audiochecker says: "this track is CDDA with probability 99%". So I am pretty sure your Audio file was not lossless... Maybe you could do the same procedure with the new, truly lossless file? I will do an ABX test myself the next days when I have the time.
I checked the spectrograms of both the version I have (not normalized) and the version on pony.fm and they look absolutely identical... one is 1057kbps in FLAC and the other 1059... but they could have different compression settings
Test with this one (non-normalized Bandcamp version)
https://drive.google.com/file/d/1Tp79tJC9VRBAqLBNdu9pzvZQctej2yjj/view?usp=sharing
Or I could probably test it...
Omg... lmao... the one I linked was to the 320 mp3 normalized version.... that would explain it...
Um... lemme make sure I tested with the lossless..
whaaaat... how could I make that silly of a mistake... omg... I thought I had normalized the lossless one... pfft... anyway, the one on Bandcamp is lossless I just have to test it
Hang on, no.... no....
Hang on... I'm so confused right now.. lemme collect my thoughts and... ugh...
Okay... the one I tested was the lossless one... just now I was looking at the spectrogram of the 384kbps Opus one so of course that cut off at 20khz and made me think it was the 320 mp3 version....
I just re-tested just to make sure with the truly lossless from Bandcamp with 320kbps Opus and the artifacting was extremely audible... didn't even need to ABX... I started the ABX test but then was immediately able to tell which one was which from the 1:45 to 1:50 mark or so...
The one I linked to here was stupidly the 320 mp3 version... you can tell since it cuts off at 20khz... but I'm 98% sure I tested with the lossless on my end.. I just linked to the wrong one on Google Drive...
Actually, I just looked at the ABX test... I'm 100% sure I tested with the lossless now.... the file names are different from the one I linked to and the one I actually listened to
This is the one I actually tested with and is truly lossless: https://drive.google.com/file/d/1jhw--FQbS31MIxd_v848MK5c6gjw3V3K/view?usp=sharing
I don't know how I could make such a silly mistake as linking to the normalized 320 mp3 one that I made a while back...
Ok, but I need the lossless normalized version to compare with the normalized opus. Your lossless FLAC is identical to the one on pony.fm, no difference in audio.
I just linked it in my last post....
It's the actual file I used to test with... I just accidentally linked to the wrong one on Google Drive
Was the normalized Opus converted from the lossy normalized FLAC or from a normalized FLAC I don't know yet? At the moment I have the lossless original FLAC, a lossy normalized FLAC and a normalized Opus file with unknown origin.
I took the FLAC put it in audacity then normalized it down 5db then re-saved as 16-bit wav then took the wav and compressed into Opus
Then I ran both the normalized wav and the Opus through an ABX test
Could you upload your narmalized wav (flac is also ok of course). Doing the testing with the lossy normalized FLAC you already uploaded would make no sense.
This is strange... Your Opus files all have a sample rate of 44.1 kHz. I always thought Opus would only support 48 kHz! When I convert a 44.1 kHz source to Opus (using fre:ac, Opus Audio Encoder v1.2.1) it has always a sample rate of 48 kHz...
Um... Should be 48khz...
You sure you're looking at one of the .opus or .ogg ones?
I can check when I get home but almost positive it's 48khz
On Tue, May 1, 2018, 12:05 PM damian101 [email protected] wrote:
This is strange... Your Opus files all have a sample rate of 44.1 kHz. I always thought Opus would only support 48 kHz! When I convert a 44.1 kHz source to Opus (using fre:ac, Opus Audio Encoder v1.2.1) it has always a sample rate of 48 kHz...
Well, it's wrong...
Audacity and foobar both say it's 48khz
It's probably stupidly looking at the "original sample rate" and displaying that
https://i.imgur.com/KpBpfD7.png
Yes, you're right, they are all 48 kHz, but why does MediaInfo show 44.1 kHz for your files and 48 kHz for mine? Aren't we using the same encode? Makes no sense... What about the lossless noramalized FLAC/WAV? Can you upload it?
As I said before, I linked it in a previous post
sigh I'll link it again...
https://drive.google.com/file/d/1jhw--FQbS31MIxd_v848MK5c6gjw3V3K/view?usp=sharing
"Yes, you're right, they are all 48 kHz, but why does MediaInfo show 44.1 kHz for your files and 48 kHz for mine?"
What do you mean? You mean you tried running a file you encoded with Opus and checked with MPC-HC to see the sample rate?