The state of lossy audio in 2022.

[Smitty2k1]

For playback of local media, FLAC all the way simply for archival purposes. If I'm taking the time to copy a CD may as well do it right. If some other format replaces MP3 in the future I'll be happy to have my lossless FLAC archive. Hard drives are cheap and audio takes up virtually no space compared to video backups.

For streaming I'm much less picky.

 

[threni]

Lately there have been a few threads complaining about the lack of fidelity for lossy audio. I figured that it may be worthwhile to better illustrate where we are actually at with a few samples. Given that our auditory memory is only a matter of seconds and most tests play out a long chunk of a song, I decided to prepare a perfect loop track from a random song. What better random track than a random song from Daft Punk's Random Access Memories? I've formatted the samples as shown below, and it will be best to play in a player capable of gapless playback like foobar2000. Make sure you go to Playback > Order > Repeat (playlist) if you'd like the playlist to loop.

I'm running the test through my focusrite 2i4 > balanced > JBL 306pmkii's and I'm having a really hard time telling them apart, even the lower bitrate tracks. I'm really curious to hear what everyone thinks. What are your thoughts?



Download samples here:

Lossy Test - Google Drive

drive.google.com

foobar2000:

foobar2000

foobar2000

www.foobar2000.org

Study comparing 192Kbit MP3 and AAC vs wav:
https://downloads.hindawi.com/journals/ijdmb/2019/8265301.pdf

Study comparing lower bitrate Opus, AAC, Ogg Vorbis and MP3 to a lossless source track:
Results of the public multiformat listening test (July 2014)

View attachment 189253

I was getting pretty sick of copying a subset of my flac collection via wifi onto my Pi Zero W so I've decided to check out lossy formats for the first time in years. I'd never heard of Opus before but it seems that 128kbps files are pretty good in terms of the size/quality trade-off. Roughly 1/7th the same of my flacs and a quick and dirty a/b test doesn't highlight any differences.

Anyone else here used Opus, done proper a/b tests etc? I'd never have considered mp3 files this small - indeed, my first experience of mp3 was encoding a bunch of my CD rips to 160kbps mp3 back in the day (15/20 years ago) and they didn't sound too hot. I went with 320kbps for some years after that before settling on flacs.

(Note: My main system will always be flac! This is a second, portable system: moOde installed on a 200gb sd card, my old HD 595 headphones and the US Apple dongle. Not interested in: a larger card/ethernet/streaming/having fewer flac files etc etc.)

 

[itz_all_about_the_music]

Does anyone here remember a Sony sponsored live event in the '80's where they played back to small-ish audiences: a) CD 16/44.1, b) MP3, c) the delta signal (i.e., CD minus MP3)? The operative in their demo was challenging the listeners to identify that (c) would make any difference in your listening enjoyment. It was interesting, and all participants got a Sony t-shirt. ;-)

 

[MKreroo]

My library is mostly WAV and FLAC, but it's mainly for archival purposes, in case I want to re-code into other formats.
An Opus version of my library is on my phone where storage is quite limited.

Personally. I can't really hear the difference.

 

[Suffolkhifinut]

Thanks for doing this. I already know the answer, at least for myself: in normal listening, high bitrate lossy (256 AAC, 320 Ogg Vorbis) is indistinguishable from lossless and hi-res. Given enough time and focused effort I might be able to pick out some minute differences in certain types of isolated high frequency sounds, but these would be irrelevant when listening to music for enjoyment.

Here are a couple of blind tests that support my findings:

Blind Test Results

The file size of an iTunes song is only a fraction of its CD counterpart because it has been compressed discarding data in the process. Is the audio quality degraded? More than a year ago I devised…

cdvsmp3.wordpress.com

High Bitrate MP3 Internet Blind Test: Part 2 - RESULTS

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

When is lossy MP3 better? Tuned into Blues Bar on internet radio, MP3 and a fairly low bit rate and it sounded authentic. In HiFI terms the reproduction was S*** yet it sounded authentic. Having been to many blues and rock clubs over the years and they sounded just like this! Put on an SACD of Robert Lockwood Junior ‘Discovery’ on Telarc, superb recording and sterile.

 

[Lambda]

I'm keeping everything as PCM/WAV (if this is the source format) or FLAC.
With "normal music" 320k is absolutely fine for me. but in an ABX test i can hear a difference of 192kbps mp3 reliably.
192khz is also absolute fine for casual listening.
But storage is cheap so i don't bother compressing.

 

[REK2575]

For me, ripping my cds to FLAC or using 'high res' streaming service like Qobuz is more about 'peace of mind' than anything else. I just feel comforted by the assurance that my digitized music is 44.1kHz/16 bit or better . But I still keep my Spotify subscription (partly for the family plan, but also because I actually like Spotify), and truth is, I can't hear a difference. I know for certain that I can't tell the difference between 44.1/16 bit and higher resolutions like 192/24 bit.

 

[Timcognito]

This will all be good when all your music goes into your watch. The thing that most study's lack is a good front end and a lot of what is tested which is popular electronic music, meaning electric guitar and bass or orchestra where there are so many players and such a broad sound stage that precise differentiation is difficult. A better test methodology involves controlling the source like a single voice or instrument, or small acoustic combo recorded with a closely miked recording by high end label like Chesky, 2L, Waterlily, MA etc. and played back on same set of well measuring headphones or speakers by subjects. In all the studies (I just skimmed the above) no effort is made to review what equipment is used by the people who could distinguish formats. These tests have too many variables to show that there may a significant difference in formats but do establish for the average listener high rate MP3 is pretty good.

 

[LeftCoastTim]

As an anecdote, I did some serious listening tests (abx) last summer, and try to find the "near transparency" point for different codecs. I hadn't done this in a decade or more (pre Opus), and I'm in my 40s.

For most typical music, casually listening, LAME -v4 (163kbps), Vorbis -q4 (128kbps), opus 96kbps were very much adequate. Below that, I wasn't entirely happy for some reason, maybe the low-pass filter became too noticeable.

With a particularly hard to encode sample, I tried to find 90% confidence abx point (meaning I was starting to make mistakes). I reached this point it at LAME -v0 (224kbps), Vorbis -q6 (192kbps), and Opus 160kbps.

At least for me, these bit rates would be "effectively transparent" for all music.

File sizes don't matter too much anymore with such large microSD cards available now. But still, if I wanted efficiency for some reason, lossy audio is more-or-less a solved problem with Opus and Vorbis.

 

[Mulder]

When it comes to ripping my own CDs, FLAC is an obvious choice for me. Not because I necessarily hear a difference, but to preserve the CD in original format if, for example, I want to make AAC files from CDs in the future. When it comes to streaming, it doesn't matter much, but at least one reason to stick with FLAC might be if you're using wireless headphones and streaming over Bluetooth

 

[ahofer]

This will all be good when all your music goes into your watch. The thing that most study's lack is a good front end and a lot of what is tested which is popular electronic music, meaning electric guitar and bass or orchestra where there are so many players and such a broad sound stage that precise differentiation is difficult. A better test methodology involves controlling the source like a single voice or instrument, or small acoustic combo recorded with a closely miked recording by high end label like Chesky, 2L, Waterlily, MA etc. and played back on same set of well measuring headphones or speakers by subjects. In all the studies (I just skimmed the above) no effort is made to review what equipment is used by the people who could distinguish formats. These tests have too many variables to show that there may a significant difference in formats but do establish for the average listener high rate MP3 is pretty good.

I don’t think you even need great recordings. Last time I tried, I could differentiate 328k mp3 from Redbook with classical orchestral music and jazz ensembles (every time, but smallish N), while I couldn’t with rock/pop studio creations (wrong more than half the time). Even so, I was paying a lot of selective attention to strings and cymbals.

 

[Aerith Gainsborough]

320 vs FLAC is very difficult to tell apart. I did it with a lot of time, concentration, unpleasantly high volume and direct A/B listening.

For me, during normal listening, everything beyond 225 kbps is transparent.

I still prefer FLAC for archival purposes. What if I want to put the music under a gaming video I made? Having a lossless version for that is very useful. Storage space is cheap, even my smartphone's got a 512GB micro SD card, so the hassle of re-encoding for mobile use is unnecessary.

I dislike streaming services.... mostly because I can never find what I want to listen to. :'D

 

[ahofer]

I wonder if there is a theoretical or audible advantage if you are going to feed your file to upsample for DSP?

 

[bkdc]

HydrogenAud.io

Good quality lossy is good enough. It’s a simple argument. Just like asking what what SNR your DAC needs to be “good enough”. There is no need to chase lossless. But having said that there is a ton of crappy lossy audio out there made by poor quality encoding. Or there is a ton of crappy original audio and it sounds crappy even in lossless due to poor mastering.

 

[digitalfrost]

I have participated in most of these listening tests and even provided samples for some.

Hydrogenaudio Listening Tests - Hydrogenaudio Knowledgebase

wiki.hydrogenaud.io

MP3 has been perceptually transparent since 2005 (at 128kbps). Specifically see

Results of the public MP3 listening test @ 128 kbps (October 2008)

Results of Multiformat at 128kbit/s Listening Test

Results of MP3 at 128kbit/s Listening Test

Results of the public multiformat listening test @ 128 kbps (December 2005)

 

[ahofer]

Some people even prefer it. Which is even more interesting.

 

[bevok]

Like others AAC 256 or MP3 320 sounds great to me, and usually issues I’ve spotted are present in the lossless version. Very rarely I hear oddities (often in mp3s knocking around my collection) which turn out to be encodings with dubious provenance (perhaps bad software, upsampled or something.
That said with lossless so easily available now the only reasons I listen to lossy are YouTube Music (used Play Music for years and I find it quite good for suggestions), listening on mobile data and of course listening via Bluetooth headphones. Given that AirPods are I believe by far the most popular audio product this probably means lossy is still very relevant as well as the fact that two of the biggest streaming services (Spotify and YouTube music) continue to only offer lossy.
Incidentally I had a couple of listening sessions using ldac last week (usually don’t bother as more convenient to have multiple Bluetooth connections on my XM4s) and given the number of dropouts I think the enthusiasm for lossless over Bluetooth is misplaced. I can see why Apple haven’t rushed to offer it on AirPods as has been predicted for the last couple of years.

 

[Deleted member 58722]

If this was a discussion about which online streaming service was best, I'd go in that direction. And I'd state they all have strengths and weaknesses... size of library... UI... But I'll keep my comments to the title line of the state of lossless.

As I said, to me it's about convenience and size of library... and suggestions. Spotify help me discover new music quite successfully. What I used to do in Tower Records, Spotify now does for me quite well. I have read between Spotify and Apple they always claim the leads for the largest library. But I'll openly state to me Apple has become a part of the evil digital empires, and I try to avoid their ecosystem.

The rest have smaller libraries so are not as effective.

The SQ is good enough for me for the purpose of what I use Spotify for. I have said this else where but I support my fav artists by actually buying their music, and not just streaming it. SQ solved.

Yiu can do this on spotify... i think not.

 

[Deleted member 58722]

Like others AAC 256 or MP3 320 sounds great to me, and usually issues I’ve spotted are present in the lossless version. Very rarely I hear oddities (often in mp3s knocking around my collection) which turn out to be encodings with dubious provenance (perhaps bad software, upsampled or something.
That said with lossless so easily available now the only reasons I listen to lossy are YouTube Music (used Play Music for years and I find it quite good for suggestions), listening on mobile data and of course listening via Bluetooth headphones. Given that AirPods are I believe by far the most popular audio product this probably means lossy is still very relevant as well as the fact that two of the biggest streaming services (Spotify and YouTube music) continue to only offer lossy.
Incidentally I had a couple of listening sessions using ldac last week (usually don’t bother as more convenient to have multiple Bluetooth connections on my XM4s) and given the number of dropouts I think the enthusiasm for lossless over Bluetooth is misplaced. I can see why Apple haven’t rushed to offer it on AirPods as has been predicted for the last couple of years.

I have an LDAC BT transmitter. Excellent sound, no dropouts.

 

[FrantzM]

I have participated in most of these listening tests and even provided samples for some.

Hydrogenaudio Listening Tests - Hydrogenaudio Knowledgebase

wiki.hydrogenaud.io

MP3 has been perceptually transparent since 2005 (at 128kbps). Specifically see

Results of the public MP3 listening test @ 128 kbps (October 2008)

Results of Multiformat at 128kbit/s Listening Test

Results of MP3 at 128kbit/s Listening Test

Results of the public multiformat listening test @ 128 kbps (December 2005)

320 vs FLAC is very difficult to tell apart. I did it with a lot of time, concentration, unpleasantly high volume and direct A/B listening.

For me, during normal listening, everything beyond 225 kbps is transparent.

I still prefer FLAC for archival purposes. What if I want to put the music under a gaming video I made? Having a lossless version for that is very useful. Storage space is cheap, even my smartphone's got a 512GB micro SD card, so the hassle of re-encoding for mobile use is unnecessary.

I dislike streaming services.... mostly because I can never find what I want to listen to. :'D

+0.89 because, I love streaming and find amost anything I want or dicover things I didn
t even know existed !!, Spotify in particular but have come to despise Apple Music because of the stupidest interface since ...

Lossy audio is one of those sobering, even humiliating things, audiophiles are bound to face ( off-topics, another is, the ongoing onslaught of commodity-level high performance audio components, with $50.oo SOTA , IEM and, audibly transparent DAC/Amp @ $9.oo).
320 kbps , mp3 removes at least 75% (!!!) of what was in the file containing the lossless signal. "removes" as in discards... Yet... Most of us can't tell, that include those who-believe-their-ears.

Wow!

Peace.