Opus Codec
Jan 11, 2018 at 5:15 AM Thread Starter Post #1 of 43

TheTrace

100+ Head-Fier
Joined
May 30, 2013
Posts
240
Likes
35
I was looking into this Codec recently in an attempt to reduce space on my phone. I already use AAC 256 vbr for my portable library and archive the FLAC files on my hard drive.

Currently I'm ripping my files to opus @ 160 kbps just to be on the safe side that it is completely transparent.

Does anyone else have any experience with the codec or thoughts to share? I'm mainly wondering if 160 is overkill or if I should have made it 192.
 
Jan 11, 2018 at 8:40 AM Post #2 of 43
160 is overkill IMO. 128 VBR is considered transparent for Opus from everything I've read over at Hydrogenaudio about the codec, howerever they say more testing is required in the link below. I did some abx testing myself when I decided to try Opus and found it gets hard for me to tell a difference between 96 to 128 bits vs. 256 AAC, 320 mp3, and flac. Some folks can go as low as 64 bits and hear no difference. After my testing, I encoded all of my flac to Opus at 128 VBR for use on my phone and Rockboxed Sansa Clip. I went with 128 to be safe, but I could have probably went 96 and never had noticed any difference. Opus literally allowed me to get my entire music collection on a 200 gig card (and it's a large collection) and still have a lot of room to add even more. As my signature says, Opus is my new favorite lossy format. I'm extremely pleased with it.

ETA: http://wiki.hydrogenaud.io/index.php?title=Opus
 
Last edited:
Jan 11, 2018 at 9:57 AM Post #3 of 43
Agree 100% with @fourrobert13 . I also run 96kbit opus and have found no obvious artifacts in a non-testing environment. I'm considering going down to 64k to avoid getting a new microSD card. I have yet to see any tracks that can be differentiated at 128k, including some of the more famous 'mp3 killers'. Give 96k a try and see what you think.
 
Jan 11, 2018 at 10:51 AM Post #4 of 43
Thank you guys, I had a feeling that it may have been overkill, it was confusing to follow because some people were talking about using 224 or 192 and at that point it seems redundant to abandon AAC. They mentioned killer samples being transparent at 160 to 192, not too sure what those samples are but when I was testing 128 that seemed fine.

I'll go down to 128 next time around when they come out with the actual release of version 1.3.
 
Jan 11, 2018 at 11:05 AM Post #5 of 43
Thank you guys, I had a feeling that it may have been overkill, it was confusing to follow because some people were talking about using 224 or 192 and at that point it seems redundant to abandon AAC. They mentioned killer samples being transparent at 160 to 192, not too sure what those samples are but when I was testing 128 that seemed fine.

I'll go down to 128 next time around when they come out with the actual release of version 1.3.

If the question is 'when should I ditch AAC for Opus', then one answer is 'why bother if you insist on using 256k'. The real action is down low. Did they post some detectable samples at 128k Opus? I saw some complaining of regressions at 32k and some ABX success at 48k.
 
Jan 11, 2018 at 11:51 AM Post #6 of 43
No I didn't see any posted examples, just the mentioning of being able to ABX 144-160 kbps on a killer sample. Although they did admit that was just them being fussy and that the music itself is more than likely okay at that level.
 
Jan 11, 2018 at 12:10 PM Post #7 of 43
I wish I would've discovered Opus sooner...lol. Could have saved me from encoding things two to three different times. Live and learn i suppose. If I could get a day where all three of my kids were quiet at the same time, I'd try some more abx testing and see how low i can go with Opus. Opus appears to be getting more and more popular too, yet not a whole lot of new DAPs support it. It's kind of funny, but I bought a DAP when I was running out of room on my phone a few years ago. Now with Opus, I've sold my DAPs and have gone back to using my phone because the DAPs i had don't support it. Granted, I had to get an app to play Opus on my phone, but $6 or whatever for Nuetron is a lot cheaper and easier than trying to find a DAP without a lot of bugs.
 
Jan 11, 2018 at 12:23 PM Post #8 of 43
No I didn't see any posted examples, just the mentioning of being able to ABX 144-160 kbps on a killer sample. Although they did admit that was just them being fussy and that the music itself is more than likely okay at that level.

Not seeing that in the thread; mind posting a link?

I wish I would've discovered Opus sooner...lol. Could have saved me from encoding things two to three different times. Live and learn i suppose. If I could get a day where all three of my kids were quiet at the same time, I'd try some more abx testing and see how low i can go with Opus. Opus appears to be getting more and more popular too, yet not a whole lot of new DAPs support it. It's kind of funny, but I bought a DAP when I was running out of room on my phone a few years ago. Now with Opus, I've sold my DAPs and have gone back to using my phone because the DAPs i had don't support it. Granted, I had to get an app to play Opus on my phone, but $6 or whatever for Nuetron is a lot cheaper and easier than trying to find a DAP without a lot of bugs.

Opus was put into an IETF standard, so hopefully support will move toward universal. I did the same thing and got Neutron and haven't regretted it: they've added several useful/important features that were hard to find outside of Rockbox (compression, crossfeed, a good param EQ).
 
Jan 11, 2018 at 12:34 PM Post #9 of 43
Opus was put into an IETF standard, so hopefully support will move toward universal. I did the same thing and got Neutron and haven't regretted it: they've added several useful/important features that were hard to find outside of Rockbox (compression, crossfeed, a good param EQ).
There's not too many apps that I found that supported it outside of Neutron. I know Android natively supports it, but Neutron is hard to beat. Either way, I'm content, but it'll be interesting to see Opus support in DAPs (other than the $300+ Android DAPs that still require an app).
 
Jan 11, 2018 at 1:07 PM Post #10 of 43
Not seeing that in the thread; mind posting a link?

Opus was put into an IETF standard, so hopefully support will move toward universal. I did the same thing and got Neutron and haven't regretted it: they've added several useful/important features that were hard to find outside of Rockbox (compression, crossfeed, a good param EQ).
I think it may have been this one, not entirely sure if I recall, forgive me.
https://hydrogenaud.io/index.php/topic,114656.25.html

This was something else I found.

https://hydrogenaud.io/index.php/topic,112253.0.html
Yeah it's lack of support initially deterred me, but I decided finally I should try it out at least
 
Last edited:
Jan 11, 2018 at 2:18 PM Post #11 of 43
I think it may have been this one, not entirely sure if I recall, forgive me.
https://hydrogenaud.io/index.php/topic,114656.25.html

This was something else I found.

https://hydrogenaud.io/index.php/topic,112253.0.html
Yeah it's lack of support initially deterred me, but I decided finally I should try it out at least

Ah ok, it wasn't on the V1.3 thread. Yeah, anything with strong harmonic content and quick attack (like a harpsichord) is something Opus can have problems with (I think the maintainers even have an example). It doesn't surprise me that applause can be problematic, as it's something 'noise-ish' that isn't pure noise and is made from summing a bunch of fast attacks. I have a fair bit of harpsichord content and haven't noticed anything (just via normal listening) at 96k, but if I did the solution would just be bump up that one track to 128k.
 
Jan 11, 2018 at 2:32 PM Post #12 of 43
Ah ok, it wasn't on the V1.3 thread. Yeah, anything with strong harmonic content and quick attack (like a harpsichord) is something Opus can have problems with (I think the maintainers even have an example). It doesn't surprise me that applause can be problematic, as it's something 'noise-ish' that isn't pure noise and is made from summing a bunch of fast attacks. I have a fair bit of harpsichord content and haven't noticed anything (just via normal listening) at 96k, but if I did the solution would just be bump up that one track to 128k.
Ah very interesting. Statistically speaking, I wonder how many tracks in existence (or to narrow it down, modern day releases) contain that content.
 
Jan 11, 2018 at 3:11 PM Post #13 of 43
I had not seen the threads linked. Interesting about the clapping. I've got a few live albums i will encode to Opus to test it out. I don't have anything that involves a harpsichord so I'm safe there...lol
 
Jan 11, 2018 at 4:12 PM Post #14 of 43
Massed strings can do the same thing as applause. I have a Sammy Davis Jr CD that has a string section that artifacts even at AAC192.
 
Jan 11, 2018 at 6:11 PM Post #15 of 43
Massed strings can do the same thing as applause. I have a Sammy Davis Jr CD that has a string section that artifacts even at AAC192.
Wow I see. If opus is completely transparent somewhere between 128 and 192 (killer samples included) then that says a lot about the codec. I'm judging based off this by the way,

http://wiki.hydrogenaud.io/index.php?title=Opus
 

Users who are viewing this thread

Back
Top