Opus Codec
Jan 15, 2018 at 8:38 PM Post #33 of 43
Jan 16, 2018 at 8:54 AM Post #34 of 43
Interesting 64kbit is the number being pointed out; wonder if the Soundclouders were eyeing that hydro thread…
Found the article link over at hydro. The discussion was was on the fence about the bit rate they were using. The consensus was they should use 96 instead of 64. Either way, it's a step in the right direction for more Opus support.
 
Mar 3, 2019 at 9:33 PM Post #35 of 43
Although this thread is a bit old I wanted to say a few things about Opus. Just discovered Opus few days ago and was curious about the quality of it and understood that it is the successor to Ogg Vorbis. So I made a little testing and Opus is definately better than mp3. So much better that I would say 128kbit Opus is better than 320kbit mp3. Spectral difference in the Opus codec seems to be much less than mp3. You don't normally hear spectral difference but for people who do sampling it's important that a sound is reproduced as identical as possible also in the spectral space, because samples can be triggered at any sampling rate. MP3 files sound very different than WAV when played half speed or double speed. Well, that's the whole idea in mp3 ofcourse to cut things from the spectra that you wouldn't hear when played back at normal speed.

So anyway, is Opus perfect? Can you hear a difference as compared with the original WAV? Yes, indeed you can hear a difference even at 192kbit rates. I don't have golden ears or perfect equipment, but I can still definately spot a difference. I knew what to look for. Piano sounds tend to be the most revealing and possibly cymbals so I came up with a song, Ana Tenga by Takagi Masakatsu, and from that a short 10 second passage of fast piano notes and cymbals played simultaneously. Yep, right there I could hear a difference. It was very very small yet there it was. The piano notes seemed to lack a little bit of detail, have a little rattling in them and also the timing of the notes was slightly off. This difference is very small and normally people wouldn't be able to tell unless they consciously knew what to listen for, and have a way to fastly trigger between the Opus and WAV versions. You can train your ears to detect this difference in the sound. It is easier to hear it when played back half speed but also I could spot it right there in normal speed. I also compared with mp3 and Opus was much better. I could tell between the three and guess it right every time after a little bit of training first. Subjectively this difference is less than 0.1%.

Still I would say Opus is very good. For all practical purposes and mobile listening it would be perfect at 128kbit rate. You wouldn't hear any difference. The fact that there is still some difference is not surprising. It would be strange if there was no difference at all because the sound has been compressed further 80% as compared to a FLAC version. Or 90% as compared to a WAV version. Some information is obviously missing.

So yeah, I would use Opus from now on at 128kbit instead of mp3. You can use it with foobar2000 player. Get the Free Encoder Pack that includes it or just install the latest version of the player. I used the latest Opus 1.3 for testing.
 
Mar 4, 2019 at 1:14 PM Post #36 of 43
Where is the threshold of transparency for Opus? I've done a lot of testing of Fraunhofer MP3, LAME MP3 and AAC, and I've found their Frau is just short of transparent at 320, LAME is transparent at 320, and AAC is transparent at 256.
 
Mar 4, 2019 at 2:23 PM Post #37 of 43
Where is the threshold of transparency for Opus? I've done a lot of testing of Fraunhofer MP3, LAME MP3 and AAC, and I've found their Frau is just short of transparent at 320, LAME is transparent at 320, and AAC is transparent at 256.
There's never a clear answer on that one. Apparently it's "pretty much" transparent at 128 and is transparent from 160 -192 kbps with the exception of some killer samples, according to this page: https://wiki.hydrogenaud.io/index.php?title=Opus

Since even before the release of 1.3 (and after it) I've observed a few people claiming here and there that there's potential transparency as low as 80-96, some just going with 128.

I can't attest to how scientific or thorough their methods were to reach that conclusion, but I've seen these claims a few times across some message boards.

I personally I do 192 for the peace of mind, my phone is 256 gb and my music accounts for a little over 110 gb so the worry of the file size isn't as important as the quality to me.
 
Last edited:
Mar 4, 2019 at 4:44 PM Post #38 of 43
I haven't found anyone who can detect AAC over 256 in a blind test. I'm most interested in the upper limit, not the lower, because that tells you where the line is for practical reasons.
 
Mar 4, 2019 at 6:07 PM Post #39 of 43
I haven't found anyone who can detect AAC over 256 in a blind test. I'm most interested in the upper limit, not the lower, because that tells you where the line is for practical reasons.
I am too, but I don't feel that there's a reason to go above 192 on opus in my opinion. I haven't seen any tests for 224 or 256 for it though. I think it's commonly agreed upon that those bitrates are way too high.
 
Mar 4, 2019 at 7:44 PM Post #40 of 43
For 99.9% of music 192 AAC is fine. I happened to find one song that was a real codec buster. That one requires 256 AAC.
 
Mar 4, 2019 at 7:46 PM Post #41 of 43
There's never a clear answer on that one. Apparently it's "pretty much" transparent at 128 and is transparent from 160 -192 kbps with the exception of some killer samples, according to this page: https://wiki.hydrogenaud.io/index.php?title=Opus

Since even before the release of 1.3 (and after it) I've observed a few people claiming here and there that there's potential transparency as low as 80-96, some just going with 128.

I can't attest to how scientific or thorough their methods were to reach that conclusion, but I've seen these claims a few times across some message boards.

I personally I do 192 for the peace of mind, my phone is 256 gb and my music accounts for a little over 110 gb so the worry of the file size isn't as important as the quality to me.

There's certainly material that is non-ABX-able at 96k (even 64k), but I had a harpsichord track that needed 128k, and I'm sure there are still killer samples at that rate. I certainly haven't found anything at 196k that was remotely worth trying to ABX. You know this, but others should note that Opus defaults to VBR and strongly discourages CBR, so straight comparison to AAC or MP3 CBR isn't totally valid.

For 99.9% of music 192 AAC is fine. I happened to find one song that was a real codec buster. That one requires 256 AAC.

Give it a try at 192 Opus and see what you hear.
 
Jan 24, 2021 at 1:54 PM Post #42 of 43
I have only recently started trying out the opus codec. My ears are old (54) and I don't use super high-end audiophile equipment. I mostly listen to music played from my gaming laptop into my headphones or Klipsch ProMedia speakers. Or from a DAP to IEM's.

I have tried ABX with opus files (160kbps) vs FLAC and I cannot tell the difference no matter how hard I try.

Amazing codec.
 
Feb 22, 2021 at 3:25 AM Post #43 of 43
Since even before the release of 1.3 (and after it) I've observed a few people claiming here and there that there's potential transparency as low as 80-96, some just going with 128.

I can't attest to how scientific or thorough their methods were to reach that conclusion, but I've seen these claims a few times across some message boards.

They don't use hard samples when testing 64 ~ 112kbps. Because I could tell 96kbps few times and transparency seems to start at ~160kbps. The same place tried claiming Musepack needs higher bitrate despite a 192kbps multi test showed it performs better than AAC, Dissed it being above 224kbps despite never complaining Lame pushing 280kbps at V5?.

And Opus can bloat on tonal/synth stuff like 144k = 205kbps.
 

Users who are viewing this thread

Back
Top