I've been doign a lot of research into different lossy codecs. Though most of my collection is ripped FLACs with a very small amount of HDTracks purchases, I need to load all my music onto myu phone now, and need to pick a lossy codec, so I can maximize my storage.
I know Opus was designed for low latency to help with streaming on the Internet, something that does not matter when playing flies on your phone local storage.
Both codecs support gapless playback
Both use Vorbis tags.
Both are supported on Android now
A much as MP3 is pretty universal, lack of native gapless playback support made me cross it off my list.
AAC was a good choice, and I am on Linux. And I need to go though some hoops with wine to get Apple's AAC encoder to work, whereas Ogg Vorbis is supported out of the box on all Linux distros.
Other than possible bitrate at which transparency is achieved, is there any other reason I would choose Vorbis over Opus or vice versa?
As far as I know, the is no equivalent to vorbisgain for Opus yet.
I'm on Linux too, and that's the only reason I haven't made the switch from Vorbis to Opus.
Well, there's a good reason to use Ogg Vorbis. What bitrate do you encode at?
What bitrate do you encode at?
It depends on the space I have left on my phone, but right now I use -q3 (112 kb/s). For other uses (PC, TV...) I use -q5 (160 kb/s).
There is nothing wrong with using Vorbis, it's fast, mature, reliable, and quite good even at low bitrates. I use AoTuV with enzo's patches (https://www.hydrogenaud.io/forums/index.php?showtopic=109766), but even stock libvorbis is very good.
Before I was using AAC-HE encoded with fdk_acc for low bitrates (which I found a bit better than Vorbis for 80-96 kb/s in my ABX tests), but I could not get gapless playback to work reliably on Android (depends on the player), and the ReplayGain implementation is not very flexible (AACGain sets a fixed gain, and you can not change it at playback time).
Compatibility is better with vorbis and at those bitrates they will have comparable quality.
Both codecs support gapless playback
That might be true in theory, but last time I checked Vorbis was still the only lossy codec that passed a problematic gapless test. Opus failed, just like MP3 and AAC.
(Though it's not easily audible and good gapless playback depends more on the audio player than the codec. For example Foobar on x86 Windows works great with any codec, but gapless playback on Android can fail even with lossless.)
loudness-scanner (https://github.com/jiixyj/loudness-scanner) works with opus files if linked against a recent enough taglib (https://github.com/taglib/taglib/releases/tag/v1.9).
ffmpeg copies all tags including the REPLAYGAIN ones. The peaks might not be correct. But I wouldn't expect that to matter (if an opus file has a peak > 1.0, the gain would probably be negative anyway).
Historically I've used Vorbis at quality 2 on my phone, but switched over to Opus a couple of months ago with the same target bitrate. I saved no space obviously, but Opus buys me a little more quality buffer over an already excellent Vorbis. I made the switch purely for subjective reasons as I perceived no issues with Vorbis and would still be happy as a clam using it today. I've observed no real change in quality with the switch (I had no real quality issues with Vorbis at this bitrate) and my gapless audio is just as gapless as it was before.
I'd suggest Vorbis if you intend to play the files from your phone on anything else. If not, Opus all the way. As support for Opus becomes more ubiquitous the Vorbis recommendation will likely go away. Right now, where latency doesn't matter, I basically consider Vorbis and Opus interchangeable for my purposes. I expect Opus support to gradually eclipse Vorbis support on devices so that is the direction I'm trending. All things being equal the only thing Vorbis has going for it today over Opus is the installed base of devices which are compatible. Once that advantage disappears I will consider Vorbis a completely deprecated codec.
Having lossless sources, I always encode to the best option the target device supports.
Historically I've used Vorbis at quality 2 on my phone, but switched over to Opus a couple of months ago with the same target bitrate.
Since battery life is an important parameter with mobile phones, can you elaborate on that regarding Vorbis vs. Opus? if memory serves me right Opus decoding performance is significantly worse than Vorbis, at least on x86.
I've honestly not noticed a difference in battery usage. Both are harder on my S3 than playing mp3 which it barely notices, but I've never run the thing down by even 25% playing Vorbis or Opus over bluetooth, even through hours of sitting in traffic both ways.
Last time I've done benchmarks on Opus (128 kbps, stereo, 48 kHz, vbr, full complexity = highest quality) decoder it was economic Moto G smartphone. The battery has lasted ridiculous amount of time. 35 hours 32 minutes. It was Opus version 1.1.
As far as I can see there is 1.1.1-rc with new ARM optimizations on https://www.opus-codec.org (https://www.opus-codec.org)
Saratoga has already reported that now Opus decoder is even faster than Vorbis on ARM9E
https://www.hydrogenaud.io/forums/index.php...st&p=909049 (https://www.hydrogenaud.io/forums/index.php?s=&showtopic=110284&view=findpost&p=909049)
As for smartphones there wasn't and there won't be any noticeable differences between LC-AAC, Vorbis and Opus because minimum "Idle" frequency is way too high 200-300 MHz (in some cases 400 MHz already and keep bumping). Too fast, too energy efficient and getting even better http://images.anandtech.com/doci/9330/a53-...curve_575px.png (http://images.anandtech.com/doci/9330/a53-average-curve_575px.png) (the graph is from this article (http://www.anandtech.com/show/9330/exynos-7420-deep-dive))
P.S. Link corrections, grammar and misc stuff.
There's also the likely resampling from 44.1 kHz to 48 kHz and possibly back. Though I have the feeling the lossy encoding itself is more likely to be audibly problematic.
Both Android and Opus handle resampling without issues http://geeknizer.com/audio-improvements-in...0-l-audiophile/ (http://geeknizer.com/audio-improvements-in-android-5-0-l-audiophile/)
As far as I can see there is 1.1.1-rc with new ARM optimizations on https://www.opus-codec.org (https://www.opus-codec.org)
Saratoga has already reported that now Opus decoder is even faster than Vorbis on ARM9E
https://www.hydrogenaud.io/forums/index.php...st&p=909049 (https://www.hydrogenaud.io/forums/index.php?s=&showtopic=110284&view=findpost&p=909049)
Interesting, thanks.
There's also the likely resampling from 44.1 kHz to 48 kHz and possibly back. Though I have the feeling the lossy encoding itself is more likely to be audibly problematic.
I'm glad somebody else mentioned this. I buy CD's. While modern re-sampling programs mentioned at HA are unlikely to add audible distortions or artifacts, the problem is I do not like to re-sample. Why Opus did not include native 44kHz support was a mystery to me (a separate discussion entirely unrelated to this topic).
That being said, I plan to get the newest Vorbis binaries and see how they compare to LAME.
I buy CD's. While modern re-sampling programs mentioned at HA are unlikely to add audible distortions or artifacts, the problem is I do not like to re-sample. Why Opus did not include native 44kHz support was a mystery to me (a separate discussion entirely unrelated to this topic).
The only answer I was able to find is this (http://www.hydrogenaud.io/forums/index.php?showtopic=97051):
Opus does not natively support 44Khz. The idea is that since 48Khz can store a 44Khz signal, this simplifies the algorithm. Then, it just stores the original format and size so that a decoder could decode to the same values, but a decoder is allowed to decode at another samplerate.
You might need the --raw-rate 44100 switch to tell it that the input is working at 44Khz so that it resamples.
Else, I believe it might be encoding directly as if it was a 48Khz signal, so playback would sound wrong (Haven't tried and I don't know how the commandline encoder works in this specific case).
You might opt to resample directly in the program feeding the stream (like foobar2000), or let opus resample. I don't know about the internal resampler's quality, but I know it's not bad.
I was not following the development of Opus from the beginning and when I get some information I was not able to find a definitive answer if Opus is the successor of Ogg Vorbis.
In many respects Ogg Opus is the successor to Ogg Vorbis, and was always intended to be. Vorbis could never fully carry the FOSS flag, so a whole new codec was developed. It has many potential improvements over Vorbis but these are most apparent in areas you probably don't care about. Quality at low bitrates is much better, latency is much better, but quality at very high bitrates is still suspect. Despite being transparent on a majority of test samples at lower bitrates than Vorbis (and virtually every other codec for that matter) it still occasionally stumbles over killer samples and this doesn't seem to be entirely cured at any bitrate. Despite having written a dedicated Opus plugin myself, it still isn't my preferred codec for music unless I'm trying to cram a lot of music onto limited space for casual listening. I can't get past knowing that every now and then I'll hear something that may (or may not) be an artefact. I know they're there lurking and it spoils my enjoyment.
Resampling from 44.1 to 48 should be the least of your worries though
In many respects Ogg Opus is the successor to Ogg Vorbis, and was always intended to be. Vorbis could never fully carry the FOSS flag, so a whole new codec was developed. It has many potential improvements over Vorbis but these are most apparent in areas you probably don't care about. Quality at low bitrates is much better, latency is much better, but quality at very high bitrates is still suspect. Despite being transparent on a majority of test samples at lower bitrates than Vorbis (and virtually every other codec for that matter) it still occasionally stumbles over killer samples and this doesn't seem to be entirely cured at any bitrate. Despite having written a dedicated Opus plugin myself, it still isn't my preferred codec for music unless I'm trying to cram a lot of music onto limited space for casual listening. I can't get past knowing that every now and then I'll hear something that may (or may not) be an artefact. I know they're there lurking and it spoils my enjoyment.
Resampling from 44.1 to 48 should be the least of your worries though
Why do you think that Vorbis could never carry the FOSS flag. Vorbis is supposed to be a patent free codec, whereas there are known patents covering Opus that Xiph.org has negotiated royalty free rights for Opus users.
The Wikipedia article on Ogg Opus:
As an open format standardized through RFC 6716, a reference implementation called opus-tools is available under the New BSD License, currently the only implementation. The reference has both fixed-point and floating-point optimizations for low- and high-end devices, with SIMD optimizations on platforms that support them. All known software patents which cover Opus are licensed under royalty-free terms.
The Wikipedia article on Ogg Vorbis:
Knowledge of Vorbis' specifications is in the public domain.[5] Concerning the specification itself, the Xiph.Org Foundation reserves the right to set the Vorbis specification and certify compliance. Its libraries are released under the revised 3-clause BSD license and its tools are released under the GNU General Public License. The libraries were originally released under the GNU Lesser General Public Licence, but a BSD license was later chosen with the endorsement of Richard Stallman.[46] The Xiph.Org Foundation states that Vorbis, like all its developments, is completely free from the licensing or patent issues raised by other proprietary formats such as MP3. Although the Xiph.Org Foundation states it has conducted a patent search that supports its claims, outside parties (notably engineers working on rival formats) have expressed doubt that Vorbis is free of patented technology.[47]
The Xiph.Org Foundation has not released an official statement on the patent status of Vorbis, pointing out that such a statement is technically impossible due to the number and scope of patents in existence and the questionable validity of many of them. Such issues can only be resolved by a court of law.
So it sounds like Vorbis does it's best to be patent free, while Opus does it's best to be best of breed without concerns for being patent free.
In many respects Ogg Opus is the successor to Ogg Vorbis, and was always intended to be. Vorbis could never fully carry the FOSS flag, so a whole new codec was developed. It has many potential improvements over Vorbis but these are most apparent in areas you probably don't care about. Quality at low bitrates is much better, latency is much better, but quality at very high bitrates is still suspect. Despite being transparent on a majority of test samples at lower bitrates than Vorbis (and virtually every other codec for that matter) it still occasionally stumbles over killer samples and this doesn't seem to be entirely cured at any bitrate. Despite having written a dedicated Opus plugin myself, it still isn't my preferred codec for music unless I'm trying to cram a lot of music onto limited space for casual listening. I can't get past knowing that every now and then I'll hear something that may (or may not) be an artefact. I know they're there lurking and it spoils my enjoyment.
Resampling from 44.1 to 48 should be the least of your worries though
I used to act the same way. I was always listening to music and thinking that if I was listening to FLAC the bass would sound better. Or something else that just kept me from enjoying the music. Then I took the 20 tracks I listened to most and made vbr 0 MP3 rips and ABXed them against their FLAC counterparts. At the end of the day I realized, I could not tell a difference on any of my most listened to songs.
I though for sure my classical music tracks would fail and I would spot the MP3. But I didn't.
I really thank God and my lucky stars that I did that, cause I was on my way down the path of tube amplifiers, vinyl LPs, oxygen free copper cables and other placebophile nonsense. At the time, I was definitely down the path of eventually buying silver cables and other unneeded stupidity.
So, now I use Ogg Vorbis and rip to ~256K VBR. Is it overkill? Probably. I'm sure I could achieve transparency at a lower bitrate. But, I really hate doing critical listening and my phone has 192 GB of storage. I know 256K is good and I know I have the space for it. So, I'm good with it, and I enjoy my music a lot more, not sitting there and waiting for the imperfections that I now know are never gonna happen.
I was always listening to music and thinking that if I was listening to FLAC the bass would sound better. Or something else that just kept me from enjoying the music. Then I took the 20 tracks I listened to most and made vbr 0 MP3 rips and ABXed them against their FLAC counterparts. At the end of the day I realized, I could not tell a difference on any of my most listened to songs.
I though for sure my classical music tracks would fail and I would spot the MP3. But I didn't.
Try doing the same thing at -V3 or even -V4 or -V5, you'll be amazed at what a relatively primitive format (MP3 compared to Ogg Vorbis or AAC, for instance) can do with very little available bitrate, when sufficiently refined in the shape of LAME.
I found that I was able to ABX -V5 against FLAC with some accuracy, -V4 was mostly guessing, and -V3 ended up being completely random guessing.
Maybe there will be a tiny artifact in some specific passage in some specific track, but I've learned to stop worrying by now :-)
I was always listening to music and thinking that if I was listening to FLAC the bass would sound better. Or something else that just kept me from enjoying the music. Then I took the 20 tracks I listened to most and made vbr 0 MP3 rips and ABXed them against their FLAC counterparts. At the end of the day I realized, I could not tell a difference on any of my most listened to songs.
I though for sure my classical music tracks would fail and I would spot the MP3. But I didn't.
Try doing the same thing at -V3 or even -V4 or -V5, you'll be amazed at what a relatively primitive format (MP3 compared to Ogg Vorbis or AAC, for instance) can do with very little available bitrate, when sufficiently refined in the shape of LAME.
I found that I was able to ABX -V5 against FLAC with some accuracy, -V4 was mostly guessing, and -V3 ended up being completely random guessing.
Maybe there will be a tiny artifact in some specific passage in some specific track, but I've learned to stop worrying by now :-)
The ABX test is very liberating. Almost everyone I talk to that claims they can ABX a FLAC vs a high bit-rate MP3 and are willing to show results will never let ME pick the tracks they ABX. That obviously have cherry picked the tracks that they know they can spot a difference on and continue to use those few tracks to justify their supposed golden ears.
I've also noticed these golden ears always ABX against MP3s. I never see them try AAC or Vorbis files. I wonder if any of the differences they are picking up would not be there if they are using a more modern codec.
There's also the likely resampling from 44.1 kHz to 48 kHz and possibly back. Though I have the feeling the lossy encoding itself is more likely to be audibly problematic.
I'm glad somebody else mentioned this. I buy CD's. While modern re-sampling programs mentioned at HA are unlikely to add audible distortions or artifacts, the problem is I do not like to re-sample. Why Opus did not include native 44kHz support was a mystery to me (a separate discussion entirely unrelated to this topic).
I don't have a reference but I remember reading it had to do with some DACs doing better with sample rates that were 12 kHz * 2^n than with 44.1 kHz.
Supposedly it doesn't matter on good hardware but with cheap DACs they don't do 44.1 kHz as well. Upsampling to 48 kHz improves audio quality on those cheap devices.
I don't know if that is true, but after reading the PONO has what sounded like a similar issue I suspect it is true.
Certainly the declared reason for choosing 48 kHz (given the decision that only one rate will be supported) is that some low quality audio devices behave poorly at 44.1 kHz.
https://wiki.xiph.org/OpusFAQ#How_do_I_use_44.1_kHz_or_some_other_sampling_rate_not_directly_supported_by_Opus.3F
Quote from xiphwiki's OpusFAQ
Note that it's generally preferable for a decoder to output at 48kHz, even when you know the original input was 44.1kHz. This is not only because you can skip resampling, but also because many cheaper audio interfaces have poor quality output for 44.1kHz.
I don't know what he means by "skipping resampling",and I didn't know cheaper audio devices have poor quality output for 44.1kHz.
First post btw,Hello HydrogenAudio!
Quote from xiphwiki's OpusFAQ
Note that it's generally preferable for a decoder to output at 48kHz, even when you know the original input was 44.1kHz. This is not only because you can skip resampling, but also because many cheaper audio interfaces have poor quality output for 44.1kHz.
I don't know what he means by "skipping resampling",
Rather than repeat the answer to this question, just take a look at the posts above yours discussing this.
and I didn't know cheaper audio devices have poor quality output for 44.1kHz.
That is referring to lower cost PCs, which frequently run at 48kHz native resampling rate. It is not true of all PCs, or even most, but for a lot of lower cost devices, 48k is widely used.
— Vorbis supports high sample rates, up to 192kHz.
— Vorbis has a highly optimized (in terms of speed) encoder Lancer (https://web.archive.org/web/20100217183320/http://homepage3.nifty.com/blacksword/).
— It is just more popular, more hardware and software support it.
— Vorbis supports high sample rates, up to 192kHz.
And ultrasonic frequencies really only matter in mastering stages, if at all, and you wouldn't be using lossy formats there.
— Vorbis has a highly optimized (in terms of speed) encoder Lancer (https://web.archive.org/web/20100217183320/http://homepage3.nifty.com/blacksword/).
I can imagine this may be useful if you're performing a lot of encoding at many times real time, say, batch encoding at speeds faster than most solid state drives can read the raw PCM samples at. Also, mainline and forks have significantly improved quality and speed since 2010, which appears to be when that page snapshot is from.
— It is just more popular, more hardware and software support it.
I will give you that, but support will happen in the future. There's already the huge Google push behind YouTube, which uses Opus everywhere as its highest quality or preferred format.
I will give you that, but support will happen in the future. There's already the huge Google push behind YouTube, which uses Opus everywhere as its highest quality or preferred format.
Not all YouTube videos have an Opus track available.
Google also made bold claims in the past about replacing h264 by WebM & VP8 everywhere and none of that happened.
— Vorbis supports high sample rates, up to 192kHz.
— Vorbis has a highly optimized (in terms of speed) encoder Lancer (https://web.archive.org/web/20100217183320/http://homepage3.nifty.com/blacksword/).
— It is just more popular, more hardware and software support it.
But will it blend?!
Google also made bold claims in the past about replacing h264 by WebM & VP8 everywhere and none of that happened.
YouTube only uses VP8 and Vorbis on old videos with a low amount of views.
I will give you that, but support will happen in the future. There's already the huge Google push behind YouTube, which uses Opus everywhere as its highest quality or preferred format.
Not all YouTube videos have an Opus track available.
Google also made bold claims in the past about replacing h264 by WebM & VP8 everywhere and none of that happened.
They actually did use VP8 for a while, but have since switched youtube over to VP9 on browsers that support it.
That might be true in theory, but last time I checked Vorbis was still the only lossy codec that passed a problematic gapless test.
Where is the test you're referring to?
...it still occasionally stumbles over killer samples and this doesn't seem to be entirely cured at any bitrate.
Can you provide such samples?
I recently ripped some CDs to Ogg Vorbis, using quality 6 (the quality recommended on their website for near cd quality and "lossless stereo). I found the mids to trebles to be over amplified. This resulted in audio characteristics such as snares and claps being disproportionately LOUD, and songs just generally sounding harsh.
Has anyone else found this with Ogg Vorbis?
Which encoder did you use?
I recently ripped some CDs to Ogg Vorbis, using quality 6 (the quality recommended on their website for near cd quality and "lossless stereo). I found the mids to trebles to be over amplified. This resulted in audio characteristics such as snares and claps being disproportionately LOUD, and songs just generally sounding harsh.
Has anyone else found this with Ogg Vorbis?
Did you try and ABX to rule out observational bias?
I will give you that, but support will happen in the future. There's already the huge Google push behind YouTube, which uses Opus everywhere as its highest quality or preferred format.
Not all YouTube videos have an Opus track available.
Google also made bold claims in the past about replacing h264 by WebM & VP8 everywhere and none of that happened.
Last time I checked, the current target is AV1 to be the de-facto standard codec for video in Youtube. However, re-encoding old videos into newer lossy codecs makes no sense, since that would only reduce quality further. It's easier to keep the old stuff and only re-encode if it is absolutely imperative (like preview-size, etc) and where quality absolutely isn't the issue.
I recently ripped some CDs to Ogg Vorbis, using quality 6 (the quality recommended on their website for near cd quality and "lossless stereo). I found the mids to trebles to be over amplified. This resulted in audio characteristics such as snares and claps being disproportionately LOUD, and songs just generally sounding harsh.
Has anyone else found this with Ogg Vorbis?
That doesn't seem to be a problem added by Vorbis. At much, it sounds as if you got your hands on a CD that has pre-emphasis, and as such, it needs to be de-emphasized first.
Last time I checked, the current target is AV1 to be the de-facto standard codec for video in Youtube. However, re-encoding old videos into newer lossy codecs makes no sense, since that would only reduce quality further. It's easier to keep the old stuff and only re-encode if it is absolutely imperative (like preview-size, etc) and where quality absolutely isn't the issue.
Google won't transcode from VP9 to AV1.
Google keeps original versions of uploaded videos on Youtube.
Last time I checked, the current target is AV1 to be the de-facto standard codec for video in Youtube. However, re-encoding old videos into newer lossy codecs makes no sense, since that would only reduce quality further. It's easier to keep the old stuff and only re-encode if it is absolutely imperative (like preview-size, etc) and where quality absolutely isn't the issue.
Google won't transcode from VP9 to AV1.
Google keeps original versions of uploaded videos on Youtube.
Isn't that what I've said?
Not sure.
What I mean is that Google can easily encode ALL videos to AV1 because they have an original sources (not VP9 encoded video which streamed on Youtube) which were uploaded by users.
Hi. I was having some trouble getting results from Opus at maximum quality, when encoding https://slugdge.bandcamp.com/track/spore-ensemble from FLAC. This track in particular has heavy use of reverb at the 5:50 mark, and Opus was just doing it wrong. Ogg wasn't much better until I used:
oggenc -q 10 --advanced-encode-option disable_coupling
the disable_coupling seemed to fix the loss in spatial effects that I was getting when not using that oggenc option. I believe it's similar to turning off "Joint Stereo"
I'm pretty happy with the resulting ogg file, 27M from a 95M FLAC. I am definitely recommending trying those options if you have an troublesome metal song that seems substandard with the Opus codec. I didn't try Wavpack.
[...] oggenc -q 10 [...]
really, q10 ? ~450kbps and p/acoustic fix for lossy? why? for what?
[...] I didn't try Wavpack.
hybride? someone has been very bad :P ... try it immediately! and don`t forget lossywav. or, if you so love lossy, musepack.
FWIW, Google considers Ogg Vorbis to be deprecated. I was reading a bug report from a while back that got closed WONTFIX with this comment.
Unfortunately, they refuse to implement proper support for Opus in the system codecs, so using it properly requires bringing in something like VLC.
One of the big sticking points with Vorbis was that it never lived up to its full potential because the reference encoder (which is unfortunately what gets packed into Linux distros like Fedora and Ubuntu, ugh) didn't merge back the quality or performance improvements of the AoTuv and Lancer patches. So you have to figure out how to use a forked encoder to use Vorbis properly.
Chris Montgomery of Xiph.org claimed that the uber conservative nature of reference is because AoTuv had a few bugs when it was last merged and he didn't want to deal with regressions.
As is usual with the Cascade of Attention Deficit Teenagers software development model, instead of improving Vorbis, they created something new and incompatible instead.
Even in Free Software, support for Opus isn't very good in some cases. Like some tagging software.
So they've created a mess of codecs and forked software, half-baked implementations, and varying degrees of compatibility all around when they could have just left things alone and worked on what already existed.
Perhaps it can be argued that at ridiculously low bitrates, Opus competes well against even AoTuv, but the fact remains that this would have been more useful in 1999 (when MP3 was widely used and didn't sound good at any reasonable bitrate) than 2019. :)
Also interesting is that the last codec listening test, which was five years ago, the only format that had a "stumble over and die" sample in the ~96 kbps realm was Apple's AAC.
I am a Vorbis user, and I will stay that way. I am happy that someone did an accurate description what I always thought of Chris Montgomery and co. I am very resentful for Monty's inactivity to merge or at least consider and also to promote and appreciate Aoyoume. As an engineer I know that hard to get the format right, yet, a real engineer also gets the most of the constraints of the format once it is out there. I find him failing in that. Starting yet another codec shows no commitment and discipline, but just working at whim. I am sure soon Opus will be left behind and something else comes.
Being at Xiph the goal is not just to pollute the world with codecs, but to marketing them, work with companies and making the format widespread. I look up to Josh Coalson as he has achieved that. I am sure he worked a lot in the background with mp3 player hw manufacturers and even with Xiph so that FLAC became so ubiquitous.
Albeit I stick with Vorbis Aotuv as it is truly gapless, high quality, easy tagging, and good support, but never consider Opus, and I was fortunate to bank on H.264 and now H.265 in video (and AAC for movie soundtracks) rather than VP8/9 and now AV1 (or whatever we call it) as Google comes across just like Monty and co. No proper specs, no working with HW companies. Just rolling out endless undocumented, half-baked stuff. No thanks.
Triza
Opus is obviously better but I still use Vorbis (lateset aotuv q4) on my Rockboxed Sansa devices because it's a lot easier on software decoding.
I am a Vorbis user, and I will stay that way. I am happy that someone did an accurate description what I always thought of Chris Montgomery and co. I am very resentful for Monty's inactivity to merge or at least consider and also to promote and appreciate Aoyoume. As an engineer I know that hard to get the format right, yet, a real engineer also gets the most of the constraints of the format once it is out there. I find him failing in that. Starting yet another codec shows no commitment and discipline, but just working at whim. I am sure soon Opus will be left behind and something else comes.
Being at Xiph the goal is not just to pollute the world with codecs, but to marketing them, work with companies and making the format widespread. I look up to Josh Coalson as he has achieved that. I am sure he worked a lot in the background with mp3 player hw manufacturers and even with Xiph so that FLAC became so ubiquitous.
Albeit I stick with Vorbis Aotuv as it is truly gapless, high quality, easy tagging, and good support, but never consider Opus, and I was fortunate to bank on H.264 and now H.265 in video (and AAC for movie soundtracks) rather than VP8/9 and now AV1 (or whatever we call it) as Google comes across just like Monty and co. No proper specs, no working with HW companies. Just rolling out endless undocumented, half-baked stuff. No thanks.
Triza
About the only codec that seems safe from being dumped whenever they want to make something "better" is FLAC, because it's lossless and mature and probably isn't going anywhere.
Unfortunately, the proliferation of video codecs is worse than it is with audio codecs. The problem is compounded by the fact that whenever the MPEG people produce something, it's entirely restricted and then the Free Software side has to figure out something that is similar. But while there have been dramatic improvements between generations of _video_ codecs, this is simply not the case with audio.
Even MP3 basically works fine now, and I would argue, is more safe from being dropped from the "MUST" support in the Android documentation than anything else, except AAC. When Google is finished with Vorbis for their purposes, I wouldn't be shocked if they dropped it from MUST or even SHOULD and then you're back to needing VLC just like you currently do with Opus. :)
As we've seen from Google's behavior, they drop about one of their own products every couple of weeks, even after millions of people depend on it. Why would anything be safe? :P
Opus is obviously better but I still use Vorbis (lateset aotuv q4) on my Rockboxed Sansa devices because it's a lot easier on software decoding.
Unfortunately, there are two problems with current Sansa devices.
1. Rockbox is obviously never being ported to them.
2. They probably couldn't play Vorbis even if it was because Sandisk cut the RAM on them to just enough to decode MP3, WMA, and AAC, and one of the sticking points with Vorbis was that it requires a lot more RAM to unpack the vector codebook that resides at the beginning of each file.
As to point #2, you could be forgiven for saying "If Vorbis only needs 320 KB of RAM to decode, how much money could Sandisk have saved by scooping the brains out of their current portable media players?".
Well, the dedicated portable media player is a dying market. Some people still want them obviously. Apple killed the iPod off even though there's a market demand for it because they can use the lack of one to upsell you into an expensive phone that you may not even need (and figured out you could get people to go along with DRM if you framed it as "streaming"). (Turns out that they hadn't given up on stealing their customers' freedom. They merely took a pause to formulate a new strategy.)
But, the market is, nonetheless, shrinking, and Sandisk probably figured they could put out a cut rate MP3 player and nobody would ever know the difference except maybe the 1 customer in 500 that cares about Vorbis.
Unfortunately, they refuse to implement proper support for Opus in the system codecs, so using it properly requires bringing in something like VLC.
Or GMMP, or Foobar2k Mobile, or... probably dozens of other third party players. I have been using Opus on Android (with different phones & Android versions) for years now.
One of the big sticking points with Vorbis was that it never lived up to its full potential because the reference encoder (which is unfortunately what gets packed into Linux distros like Fedora and Ubuntu, ugh) didn't merge back the quality or performance improvements of the AoTuv and Lancer patches. So you have to figure out how to use a forked encoder to use Vorbis properly.
You make it sound (pun unintended) like the stock libvorbis encoder is bad. It is not. It is good and mature.
If I remember correctly, AoTuV is only marginally better, at low bitrates, and does that in a large part by increasing bitrates significantly for difficult samples.
As is usual with the Cascade of Attention Deficit Teenagers software development model, instead of improving Vorbis, they created something new and incompatible instead.
Yeah of course, that is totally how it went.
CELT & co. and all the work on low latency and speech encoding is also fake news I guess?
Even in Free Software, support for Opus isn't very good in some cases. Like some tagging software.
The tag format is exactly the same between Vorbis and Opus (VorbisComment). With many popular tagging libraries, like Mutagen, the code path is the same, and the program does not need to know if its Vorbis or Opus.
At this point tagging programs that don't support Opus are likely abandoned anyway.
So they've created a mess of codecs and forked software, half-baked implementations, and varying degrees of compatibility all around
Please be specific, this nothing but FUD, not backup by facts.
when they could have just left things alone and worked on what already existed.
I don't think the improvements brought by Opus (quality and feature wise), could have been implemented in Vorbis without breaking the format specification.
PS : I am not related with Xiph in any way, and I don't think everything they do is perfect, both for the Vorbis maintenance, and the Opus development, but overall I appreciate their work.
and does that in a large part by increasing bitrates significantly for difficult samples.
A good encoder is an encoder that knows when it's necessary to increase bitrate, no?
True, but in my experience it increases the average bitrate for most files in my collection, and I doubt every single one of them is a "killer" sample.
Opus is obviously better but I still use Vorbis (lateset aotuv q4) on my Rockboxed Sansa devices because it's a lot easier on software decoding.
Unfortunately, there are two problems with current Sansa devices.
Sandisk is basically out of the music player business. They're just reselling some random OEM's product in the USA with their label on it. You get whatever that random company was selling.
2. They probably couldn't play Vorbis even if it was because Sandisk cut the RAM on them to just enough to decode MP3, WMA, and AAC, and one of the sticking points with Vorbis was that it requires a lot more RAM to unpack the vector codebook that resides at the beginning of each file.
Since I spent a lot of time working on the memory usage of all of these decoders, I feel like I should chime in :) Vorbis memory usage is not dramatically different than AAC. They're both large, and they both need a lot of memory up front. This is why the Sandisk players with limited memory can only decode AAC files up to 10 minutes long or something like. Vorbis support would also be possible, but those players have simple 16 bit DSP cores for decode rather than real CPUs, and the DSP core vendor only provides a few decoders since they're not meant to be general purpose devices.
More info: https://www.rockbox.org/wiki/CodecMemoryUsage