HydrogenAudio

Lossy Audio Compression => Opus => Topic started by: shadowking on 2019-07-20 10:58:59

Title: Opus severe issues
Post by: shadowking on 2019-07-20 10:58:59
Upload:

https://hydrogenaud.io/index.php/topic,117915.new.html#new
Title: Re: Opus severe issues
Post by: LithosZA on 2019-07-20 11:09:20
Sounds about normal for a CVBR 32Kbps encode. What issues are you experiencing?
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-20 11:55:06
64k cvbr is also sub par even without abx.  What do you mean what issues ? Theres no need to even abx IMO
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-20 13:52:51
64k version
https://hydrogenaud.io/index.php?action=dlattach;topic=117917.0;attach=15157
Title: Re: Opus severe issues
Post by: magicgoose on 2019-07-20 15:59:39
TOS #8 violation, no?
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-20 16:15:42
TOS #8 violation, no?

What are you even doing here ? I posted these for Opus Developers and experts to check. And NO  - I DID NOT CLAIM CODEC A IS BETTER THAN B OR VICE VERSA. IMO , I DONT NEED TO ABX IT TO HEAR IT.  Others may need to.
Title: Re: Opus severe issues
Post by: m14u on 2019-07-20 17:26:23
hear what? lossy 32/64kbps sounds like not a cd-quality? thats NEWS!!!
for retreat-
someone hacks your account...
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-20 17:38:42
hear what? lossy 32/64kbps sounds like not a cd-quality? thats NEWS!!!
for retreat-
someone hacks your account...

1- Yes according to some 'studies' , personal tests and the mighty HA WIKI  it can come close esp @ 64k.  2 - As i said its not about CD quality per say but rather how gracefully degradation occurs. You can have audio not cd quality but not annoying, Yet another @ CD quality but annoying if trouble spot(s) occurs.

3- When encoding this sample with HE-AAC / vorbis the result is clearly more pleasing to my ears.
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-20 18:06:42
4 - The notion that one should not post  / mention audio defects  @ less than tranparent bitrates is nothing more than politicaly correct bullshit . esp since the main attraction of such codecs is low bitrate use.
Title: Re: Opus severe issues
Post by: fabiorug on 2019-07-20 18:44:57
Is a good sample, not problematic
Jm Valin (I don't know who develops opus apart him) is optimizing most for 24 - 48 kbps -96 kbps (I used today build libopus 20-07-2019)  ffmpeg opus I don't blind tested it  but I think for now it's the same.
At other bitrates I hear flanging in the drums and cymbals, even at 128 kbps.
64 kbps is surpassed. Jm Valin said in this site that he is optimizing more at bitrates 24 - 48 kbps.
Title: Re: Opus severe issues
Post by: magicgoose on 2019-07-21 10:16:34
At other bitrates I hear flanging in the drums and cymbals, even at 128 kbps.
I'd like to ask for ABX results where you can detect something like that for Opus@128 kbps
Title: Re: Opus severe issues
Post by: fabiorug on 2019-07-21 11:26:44
Yesterday I ABXed it but he's expanding for other bitrates. If you tried latest libopus build you'd know that there are lower bitrate available with CELT only minimum 15 kbps but it is trasparent at about 34 kbps (about double the bitrate).
I haven't listened to an audiophile track (it's EDM) and I don't know where to find it.
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-21 12:12:47
I decided to test 128k  - with PC speakers only. Half way through the 1st trial i decided to focus on the background noise @ 2sec mark and it becomes more obvious as its not clean. It showed through the 2nd trial.

foo_abx 1.3.4 report
foobar2000 v1.4.3
2019/07/21 21:05:14

File A: G:\Hdd-2\music\abx tests\lame\Richmond Rd Revisited.flac-00.00.00.000-00.00.24.057.flac
File B: G:\Hdd-2\music\abx tests\lame\Richmond Rd Revisited.opus

21:05:14 : Test started.
21:05:36 : 00/01  100.0%
21:05:50 : 00/02  100.0%
21:06:06 : 01/03  87.5%
21:06:16 : 02/04  68.8%
21:06:40 : 03/05  50.0%
21:06:46 : Trial reset.
21:06:59 : 01/01  50.0%
21:07:11 : 02/02  25.0%
21:07:19 : 03/03  12.5%
21:07:57 : 04/04  6.3%
21:08:39 : 05/05  3.1%
21:09:27 : Test finished.

 ----------
Total: 8/10 (5.5%)
Title: Re: Opus severe issues
Post by: fabiorug on 2019-07-21 12:15:41
I am not audio engineer and I don't know which version of Opus do you refer.
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-21 12:25:55
I am not audio engineer and I don't know which version of Opus do you refer.

libopus 1.3-rc2, libopusenc 0.2
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-21 13:38:09
I 've been playing with several other samples; angels falls 1st, mandylion_sm @ 32..48..64 and came to similar disappointments. However, Getting rid --cvbr yielded a big improvement -except- the richmond road sample that become worse with default VBR. The other trouble is the female speech sample I encoded @32k . It produces unpleasant artifacts . However, 16..24k sounded much better but why ?
Title: Re: Opus severe issues
Post by: magicgoose on 2019-07-21 13:46:10
log without checksums and attached files... not convinced
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-21 13:55:27
log without checksums and attached files... not convinced

You don't get it . This is for developers and these are non-cd quality bitrates so anyone should be able to hear it. I don't care much for 128k or even 96k currently. The issue is the type of error produced and its character. I may upload the samples for devs or others interested but not gonna bother at all with checksums etc.
Title: Re: Opus severe issues
Post by: shadowking on 2019-07-21 14:06:02
female speech:
http://www.rarewares.org/test_samples/female_speech.wv
encode it to 32k
 listen to the word 'administer' and all sibilants.
Title: Re: Opus severe issues
Post by: magicgoose on 2019-07-22 08:44:50
log without checksums and attached files... not convinced

You don't get it . This is for developers and these are non-cd quality bitrates so anyone should be able to hear it. I don't care much for 128k or even 96k currently. The issue is the type of error produced and its character. I may upload the samples for devs or others interested but not gonna bother at all with checksums etc.
So you're also implying 128k isn't "cd quality bitrate"?
Title: Re: Opus severe issues
Post by: Shinichi on 2019-07-24 10:52:27
I decided to test 128k  - with PC speakers only. Half way through the 1st trial i decided to focus on the background noise @ 2sec mark and it becomes more obvious as its not clean. It showed through the 2nd trial.

21:09:27 : Test finished.

 ----------
Total: 8/10 (5.5%)

Very interesting results. I thought that Opus at this bitrate should be transparent in most cases...
Can you detect these issues with AAC (Nero) / Vorbis @128k?
Title: Re: Opus severe issues
Post by: amariami on 2019-07-29 19:57:56
I decided to test 128k  - with PC speakers only. Half way through the 1st trial i decided to focus on the background noise @ 2sec mark and it becomes more obvious as its not clean. It showed through the 2nd trial.

foo_abx 1.3.4 report
foobar2000 v1.4.3
2019/07/21 21:05:14

File A: G:\Hdd-2\music\abx tests\lame\Richmond Rd Revisited.flac-00.00.00.000-00.00.24.057.flac
File B: G:\Hdd-2\music\abx tests\lame\Richmond Rd Revisited.opus

21:05:14 : Test started.
21:05:36 : 00/01  100.0%
21:05:50 : 00/02  100.0%
21:06:06 : 01/03  87.5%
21:06:16 : 02/04  68.8%
21:06:40 : 03/05  50.0%
21:06:46 : Trial reset.
21:06:59 : 01/01  50.0%
21:07:11 : 02/02  25.0%
21:07:19 : 03/03  12.5%
21:07:57 : 04/04  6.3%
21:08:39 : 05/05  3.1%
21:09:27 : Test finished.

 ----------
Total: 8/10 (5.5%)
Hmmm could you tested on 160k ?