Let's talk about lossy audio codecs. I did a simple experiment (encoding/decoding a WAV track on tmpfd. ffmpeg 3.3.2, opus 1.2.1, vorbis 1.3.5, lame 3.99.5 and native ffmpeg aac codec) and got these.
So, which one do you use? While Opus is clearly the leader in quality:size ratio, I still use Vorbis because this decoding speed actually matter on a Clip+.
Codec │ Encoding │ Decoding──────────────┼──────────┼──────────libopus 128k │ 3.181000 │ 0.991000──────────────┼──────────┼──────────libvorbis q5 │ 4.990000 │ 0.590000──────────────┼──────────┼──────────libmp3lame q0 │ 5.778000 │ 0.975000──────────────┼──────────┼──────────aac q0 │ 8.277000 │ 0.360000──────────────┼──────────┼──────────aac 160k │ 8.504000 │ 0.356000
Lossy audio codecs
I use opus at work in our VoIP app as it's prrtty great for that, but save audio in vorbis. Opus is kinda a weird thing to save audio in as under the hood as it flips between two codings and all the FEC stuff is wasted.
I think it switches to full CELT beyond 32kbps.
Vorbis never took off because it was way more CPU-intensive than mp3 without an ASIC. Opus is designed to be more efficient to decode than either. I think you fucked up your measurements.
Vorbis is huge commercially, most games and apps ship their audio this way to avoid the licensing Jew. It never took off with consumers as almost all their music is pirated anyway so they don't care about fees.
Who cares about CPU usage? The Vorbis usage is negligible in any machine from the last 15 years.
Try reading the OP?
Are you retarded?
IT'S FRIDAY NIGHT AND I HAVE FUCK ALL ELSE TO DO LOL
#!/usr/bin/env perl6use Bench;my Bench $b .= new;my @ffmpeg = ;sub enc($name, @params) { $name => sub { run(@ffmpeg, «-i input.wav», @params, "output.$name") }}sub dec($name, @params) { $name => sub { run(@ffmpeg, «-i "output.{$name.split('.').tail}" -f null /dev/null») }}my @enc = 'ff.opus' => , 'opus' => , 'ff.ogg' => , 'ogg' => , 'mp3' => , 'vbr.aac' => , 'cbr.aac' => , 'fdk.aac' => ;my @dec = 'ff.opus' => , 'opus' => , 'ff.ogg' => , 'ogg' => , 'mp3' => , 'ff.aac' => , 'fdk.aac' => ;$b.timethese(10, @enc.map({ enc(.key, .value) }).list);$b.timethese(10, @dec.map({ dec(.key, .value) }).list);
Benchmark:Timing 10 iterations of cbr.aac, fdk.aac, ff.ogg, ff.opus, mp3, ogg, opus, vbr.aac... cbr.aac: 125.2251 wallclock secs @ 0.0799/s (n=10) fdk.aac: 24.6223 wallclock secs @ 0.4061/s (n=10) ff.ogg: 0.3398 wallclock secs @ 29.4305/s (n=10) (warning: too few iterations for a reliable count) ff.opus: 0.3436 wallclock secs @ 29.1000/s (n=10) (warning: too few iterations for a reliable count) mp3: 23.1855 wallclock secs @ 0.4313/s (n=10) ogg: 20.0175 wallclock secs @ 0.4996/s (n=10) opus: 16.6672 wallclock secs @ 0.6000/s (n=10) vbr.aac: 36.8352 wallclock secs @ 0.2715/s (n=10)Benchmark:Timing 10 iterations of fdk.aac, ff.aac, ff.ogg, ff.opus, mp3, ogg, opus... fdk.aac: 0.2254 wallclock secs @ 44.3630/s (n=10) (warning: too few iterations for a reliable count) ff.aac: 0.1944 wallclock secs @ 51.4325/s (n=10) (warning: too few iterations for a reliable count) ff.ogg: 3.4461 wallclock secs @ 2.9018/s (n=10) ff.opus: 6.6929 wallclock secs @ 1.4941/s (n=10) mp3: 3.6070 wallclock secs @ 2.7724/s (n=10) ogg: 3.5331 wallclock secs @ 2.8303/s (n=10) opus: 6.5571 wallclock secs @ 1.5251/s (n=10)
$ du -sh input.wav9.8M input.wav
tl;dr go violate patents
Ogg isn't that bad tbh. Do some ABX testing. Most people here think they have special snowflake ears but couldn't tell the difference between 320kbps MP3 and FLAC.
Man, if only ffmpeg's AAC encoder was better.
for portable use and high quality, (still) nothing beats musepack.
MP3 chokes on certain sounds at any bitrate. Musepack doesn't, and even when it has some problems, it's never as annoying and nasty.
Also, CBR encoding of audio is a complete waste of space. It's completely pointless in 2017 and it was pointless since very long time ago.
Horseshit, Opus is even slower to decode. Still, it's the king for achieving "acceptable" quality at extremely low bitrates.
The mpc worship must stop, m8. It was good a long time ago, when mp3 didn't have Lame and Vorbis was only starting. As for speed, with musepack-tools-465, I get encoding 4.053 and decoding 0.601 (I'm the OP).
So why use it over Vorbis? You get retarded APE tags and the inability to use ffmpeg to pass metadata painlessly. Plus, with Vorbis, I can just use the FLAC replaygain tags without having to recalculate it.
Honestly, only Vorbis and Opus are worth using. If AAC had a way to convert replaygain info to its own format automatically, I'd think about it.
Well, if you have bad hearing then the lossy format you'll use hardly matters.
Vorbis WAS tied with Musepack in most listening tests. It's now over if you know how it developed with since then.
I've met one of those faggots IRL. Said he could hear high HZ sounds. In reality he was way below average and would stop hearing at around 17khz.
meanwhile I can hear the electron gun of a CRT and it's really fucking annoying
What's your point?
because it's at even lower frequency, 15625hz, dumbass
user's point is probably that Vorbis is less CPU intensive than MP3.
Also, CPU usage counts because battery life does.
If that was the case everyone would hear the constant and teeth grinding buzzing that I hear whenever a CRT is on.
But that's wrong. Vorbis is basically better than mp3 in every way. MP3 took off because it was the first usable lossy audio codec and in the 90s pirates started releasing music CDs in mp3 because of the extreme bandwidth savings and high quality.
I use .mp3 for lossy because I'm not a hipster faggot. .flac for lossless, same reason.
Opus niggers are like those FOSS evangelists that go around using Gimp and pretending it's better than Photoshop.
...
Look at the bright side, you can now set your sound devices to 32khz and enjoy lower cpu usage.
Isn't mp3 not licensed any more?
The frequency is 15k. Do a recording yourself if you don't believe the other user.
The reason other people can't hear it is because most people are retarded about protecting their hearing. Or they're older.
What is your idea of average or above average hearing? I can hear up to 18khz according to this.
audiocheck.net
Nobody says Gimp is better than Photoshop. We always say that anything you can achieve in Photoshop can be achieved in Gimp. We always say that Gimp is always free unlike Photoshop.
Yeah, if you use the fucking pen tool as a single pixel, set your color by RGB value and draw the fucking thing pixel by pixel
You're doing it wrong if you use an image manipulator as a drawing program. Use a drawing program if you want to draw.
Most normies are like "everybody is okay with it so I must be okay with it".
Of course it's a horrid sound.
Shockingly, 3GP is bretty good for voice compression at low bitrates. I know because I download big audiobooks off YouTube and it's very small and sounds much, much better than MP3 at that size.
it's a container format, dumbass
you are probably having AAC or AMR inside it.
Good job
Does opus have replaygain support yet?
Last time I checked, it was a clusterfuck.
replaygain goes in the container format, dumbass
what made you think so?
I don't have to look up trivial shit in wikipedia.
nothing stops you from shoving opus into an already widely understood container format like OGG.
There is absolutely no reason to be using anything but Opus for lossy audio storage these days unless your shitty audio file player has problems with it. The jury is still out on the best video codec, but Opus reigns supreme among audio.
A few years back when I last checked, GIMP Animation Pack actually does manage to achieve more efficient delta frame cutting for animated GIFs than whatever Photoshop uses. But GIF is quickly becoming a format of the past. GIMP's impressive delta frame cutting could be useful for APNGs if someone would get that APNG plugin functional enough to use. GAP also needs a rewrite to specify the far smaller frame display times that APNG supports.
Many Android phones are still "shitty audio players," sadly. My Samsung Edge 7 can play Opus files natively, but it can't seek them, so when I stop playing or close an application it's back to the beginning.
Opus, as any transform-based codec, deals very inefficiently with transients and for some samples it just can't encode them faithfully without wasting a lot of bitrate. This method simply isn't optimal for "transparent" quality level. Yeah, many people don't hear the difference, but not all…
Musepack on the other hand can encode everything faithfully without such problems, because it works on totally different principle.
you got what you asked for.
Nexus 5X FTW
also, use a proper audio player application, the stock one is probably kill.
MPC is transparent at 160-180, Opus is at 128. The inconvenients of subband (bitrate wise) are simply larger than MDCT.
Anyway, since the beginning, you're praising mpc, but do you have any proof of what you're saying? Like recent listening test with a big sample.
All Android distros use the same audio libraries and every Android audio player except VLC (which is shit) uses those native libraries for playback you tween retard.
Just save as MNG and cuckify it in ffmpeg :^)
Except when it isn't.
If it was true then indeed other codecs are btfo.
at least ddb2 doesn't (it's non free though)
wiki.hydrogenaud.io
ctrl+f "killer"
you can search the web for these samples yourself, it's not that hard.
Might aswell use it unless you really care about battery in which case go for AAC which is better than mp3 and lighter.
in terms of battery Musepack is the king (at least if others are not using some special snowflake hardware decoders) so might as well use it
How does FLAC compare?
The title of the thread is 'lossy audio codecs'. This means that the exact audio data is not preserved after encoding; some of it is thrown away in order to minimise file size. FLAC (Free Lossless Audio Codec) is the opposite of lossy, lossless. The audio signal you get from a flac encode of, say, a CD, is exactly the same as the original. No data is thrown away, but compression is used to minimise the space this data takes up.
So it's not very useful to compare FLAC with Opus, mp3 or vorbis. However, flac is quite mature and encoding is pretty fast. In my experience, flac encoding an entire CD takes a minute or less and reduces the size by 40% to 50%, depending on the type of music.
autism
no u
But it's not. See
Are you sure that test was done on the same hardware arch that you're gonna use in your portable player?
Test it on ARM first, dude. Then make conclusions.
And speed on x86 hardly matters in 2017 as long as it's not too slow. Because nobody wants to use x86 anywhere except desktops/laptops/servers, and they're powerful enough so that decoding of any of these formats in realtime will take
Well, do it then m8. I could do it, but you're the one with incredible claims.
That's doesn't make the comparison itself invalid. The only thing that could matter is the need of an FPU, vorbis has tremor.
But I'll try with QEMU maybe.
Which does better at encoding multichannel audio?
multichannel is kill.
music doesn't need more than 2 channels, because we have 2 ears. just some audio engineers are too impotent to do a good stereo mix, it's a lot easier to just throw shit around in 4-6 channels and be done. but the end result is shit.
Fuck the people who mix this garbage.
Bingo. If you only listen to music on a machine with a physical keyboard, there is no reason to avoid opus. The next problem is searching for Idolmaster FLACs with full tags.
inb4 blackberry
mixing is a art
I don't need to prove shit to other people. If you don't want to listen with superior quality/size, then that's your choice.
Bonus: since money was put in all those other codecs, they have the NEON codepaths you want.
Multichannel could be done well (cf ambisonics), but what we have currently is a stupid gimmick, indeed.
This shit here pisses me off to no end.
I want to hear what people say and I don't want to lose my hearing by the time I'm 40
How does libfdk_aac compare to aac?
How does libvorbis compare to vorbis?
How does libopus compare to opus?
I'm talking decoding speed, encoding audio takes too little to matter.
I have that sometimes too on shit mixes but if you get it a lot, try bumping your center channel up a dB or two since most dialog is there.
nodoby said that.
these codecs just use different approach and have a different goal (achieve ok quality at the smallest possible bitrate) vs musepack goal (achieve completely fucking perfect quality at smallest possible bitrate).
it turned out that these goals have different optimal methods (until someone founds something else which is optimal at all quality levels)
20 years ago doesn't mean something important which could benefit musepack has been invented since then, because it already has very advanced psy model (in other words, it really knows how to distribute distortion so that it'll be masked), and everything else can be improved for the general case only by small amounts (even better entropy coding? probably will win at most 5-10%).
good headphones + binaural stereo trump all multichannel and is more efficient.
the only real point for multichannel is that some mixes tend to isolate vocals or some instruments, so it can be edited out if you want to help your gf with material for singing for example.
/founds/finds/
if the sound quality is not absolutely important for you, then this shit is solved by a dynamic compressor DSP.
just need to pick the right threshold and ratio for each movie.
Opus is superior to vorbis at any bitrate.
Here is some graph for opus 1.1: people.xiph.org
In opus 1.2 it is enabling CELT (and stereo too) even more aggressively.
opus beats the shit out of musepack
Whatever hearing you have, it always is best to choose opus.
If you have bad hearing, then opus will allow you to use 48kbps instead of 128kbps cbr mp3. If you have better hearing opus will allow 64-128kbps instead of mp3 vbr ~225kbps (or more). You always win with OPUS. It's the ultimate choice.
good goy
Not at all. Difference is Gimp is a piece of useless shit, whereas Opus is superior to any competition. Opus is probably the only good thing that FOSS niggers did in their history.
linux is piece of shit, gimp is shit, everything FOSS is shit, but opus is good. I would rate Opus project 4/5, one star is taken away because it doesn't have a GUI and I need to use 3rd party GUI.
No you cannot achieve. And also, how you achieve things is extremely important.
Photoshop is also free, they sell it for 0$ at pirate bay
Even if I wanted to use free shit, there are better freewares than gimp
do you also use windows media player on your computer?
so what? opus is transparent at 95%+ of songs at 96-128kbps
at 160kbps is transparent at 98%+ songs
it's simply not efficient to use stupid flac or musepack at way bigger bitrate because you will use few times more space and gain average of 1% in quality.
why not buying machine that can play opus? you can buy rockbox
wiki.xiph.org
m8, I've used mpc --standard for a long time and it was (still is) indeed good. But believe me when I say that it's obsolete for music.
DCT type methods have indeed their killer samples, but not musical ones; Vorbis after the aotuv changes were merged is the best replacement.
Test yourself ffmpeg is always faster for decoding.
Yeah, you're right. But imagine having one channel per instrument, thus being able to modify the mixing yourself.
How are you getting these numbers? I'd like to reproduce this.
why? battery is enough to listen for hours. if you listen more (in travel) you should have powerbank and charger anyway
So you would bring a powerbank and a charger.. what's your point? Is it; don't use vorbis beause you can just bring a charger and a powerbank? Thanks for being stupid, the internet appreciates it.
there's a different name for that: multitracks
Who is to decide what is musical and what is not?
of course, the authors writing about the software they wrote.
Your brain pulls a huge amount of info out of how sound is attenuated by your body. You don't have your own research team building a computer model of your head and so stereo will continue to be massively inferior to mulichannel. Saying otherwise is fox & grapes shit like "you can't see more than 60 fps".
nope, this is completely different.
your analogy is closer to "one can't hear higher than 11kHz", which is bs of course.
the more correct one would be "you can't distinguish colors on a scale with more than 3 dimensions" and it's true because there are only 3 kinds of color+light sensitive cells in the human eye (well it's 4 dimensions for some rare women apparently, but you should get the idea)
The point is that there's a huge difference between the two that people without the hardware claim doesn't exist because they don't have it.
what if it's in a shitty mp3 player? to awnser your question chink manufacturers care.
I'll trust them over you (or anybody). Seeing their history (flac, vorbis, speex), they obiously know their shit.
Now if you have measurements or anything substantial, I'll gladly take it.
what is best encoding for ultra low filesize? quality is of little importance.
Read the thread, nigger.
libvorbis fails to encode if you go lower than 45kilobit/s
libopus will do 7kbit/s if you try to go lower than 7kbit/s. If you try 1kbit/s it'll produce a mute 1kbit/s file instead. opus will also not do lower than 7kbit/s
libmp3lame will do 32kbit/s if you try to set the bitrate lower than 32kbit/s
libaac_fdk will do 12kbit/s if you try to set the bitrate lower than 12kbit/s
aac will do 24kbit/s if you try to set the bitrate lower than 24kbit/s
libopus sounds the best of all and it also lets you set the bitrate the lowest out of the ffmpeg encoders I tested. Here's a song at 7kbit/s opus.
get a load of this faggot without AoTuV
But user-kun AoTuV will only do 48kbit/s.
Bad Opus support is a current universal problem on Android phones you dumb nigger. Most players use the Android libraries to play music and the ones that don't are janky freetard shit like VLC.
thanks, with vorbis at 45kb/s variable bitrate my music sounds great and takes up almost no space!
Mainline libvorbis has merged with aotuv, user.
Use opus at the same bitrate instead for better quality.
AMR-NB