Why do some YouTube videos sound better than streaming music files ?

[tmtomh]

YT video soundtracks are lower in quality than lossless, or even high-bitrate lossy, streaming services.

However, like @Pearljam5000 I have periodically heard YT versions that I felt sounded good. And every single time I downloaded/extracted the audio from the YT video so I could load it into my computer-based music server and play it on my main stereo setup, the result has been disappointing. So I have just assumed that the point @Multicore makes above is actually a real thing, at least sometimes: listening while watching something seems to produce a different perception of the audio than purely listening.

I've experienced the same thing with Led Zeppelin concert bootlegs when some old archival 8 or 16mm video footage is found and someone syncs the existing audience recording to that video and uploads it to YT - while watching it, it "sounds" better than just listening to the audience recording alone.

 

[danadam]

Doesn't YouTiube use sonic compression here and there?

Not that I know of.

There's the harmless volume normalization. And I guess that in the past year or so, when they were testing "Stable volume", it could have happened that some music video was played with dynamic range compression enabled. But in general it seems to be sorted out now.

 

[Multicore]

Right click on the video, choose Stats for nerds, look up the audio format ID here

Youtube Format IDs

Youtube Format IDs. GitHub Gist: instantly share code, notes, and snippets.

gist.github.com

 

[Multicore]

Not that I know of.

There's the harmless volume normalization. And I guess that in the past year or so, when they were testing "Stable volume", it could have happened that some music video was played with dynamic range compression enabled. But in general it seems to be sorted out now.

Specifically, and iiuc, YT computes loudness for the whole uploaded video and if it exceeds -14 dB LUFS then it applies negative gain equal to the amount the upload exceeds -14. If the upload doesn't exceed -14 then there's no processing beyond the transcoding. Or so I have read.

Stats for nerds shows how much such normalization has been applied, for example, what I'm listening to right now was 4.8dB over.

Volume / Normalized: 100% / 58% (content loudness 4.8dB)

 

[danadam]

Specifically, and iiuc, YT computes loudness for the whole uploaded video and if it exceeds -14 dB LUFS then it applies negative gain equal to the amount the upload exceeds -14. If the upload doesn't exceed -14 then there's no processing beyond the transcoding. Or so I have read.

Yes, that's correct. I think it's also worth mentioning that the gain is not "baked" in the stream but only applied by the player during playback.

 

[Multicore]

I've experienced the same thing with Led Zeppelin concert bootlegs when some old archival 8 or 16mm video footage is found and someone syncs the existing audience recording to that video and uploads it to YT - while watching it, it "sounds" better than just listening to the audience recording alone.

Right! I love this VDGG video, it's thrilling and sounds awesome ... until you listen to it without video and compare it with the album version or the new stereo mix 2021. Same thing watching Focus play Hocus Pocus on Midnight Special.

 

[Chr1]

I wonder if some YT tracks are "tweaked", like they are sometimes in the soundtracks for movies. I have definitely heard tracks used in movies that sound better than the originals. They have been EQ'd, compressed and otherwise manipulated to sound better than the original recording. Pretty sure it's been confirmed here in another thread somewhere.
Could be the same with some YT content...

 

[restorer-john]

I wonder if some YT tracks are "tweaked", like they are sometimes in the soundtracks for movies. I have definitely heard tracks used in movies that sound better than the originals.

Definitely true about tracks used in movies. I think it was Top Gun Maverick where I last heard a large difference in one of the tracks. Might have been Loggins' Danger Zone?

Sounded a lot bigger and fatter IIRC than the original.

In 2018, Loggins told TMZ that he was having discussions with the film's lead actor Tom Cruise about having a new version of the song featured in the then-upcoming film Top Gun: Maverick.[12] Ultimately the original recording was used instead. Loggins stated that Cruise wanted to invoke the same feelings listening to the song as with the original Top Gun.

 

[Chr1]

Sounds about right. Reckon that it is probably more likely than not, these days. Quite likely that the sound designers used for big budget films nowadays see it as just another part of the job. Probably done by different people though. I doubt if Hans Zimmer and the likes spend their time remixing dubious recordings from the 80s.

 

[goat76]

Many times for movies, the audio engineers definitely got their hands on the multi-tracks of the music (or perhaps they just used a Dolby Atmos version of the track?), but I often find it to sound fantastic and many times way better sounding than the original stereo track. The envelopment with some of the instrumentation spread out to multiple speakers in a surround system, and sometimes seemingly with a way less dynamically compressed/limited sound it can sound very good.

 

[Enstip]

The OP’s line of questioning has been niggling me too.

My angle is slightly different though - because I’m comparing live performance with album version. I’ll give an example. The YouTube video titled “David Gilmour - Between Two Points (with Romany Gilmour) [Tour Rehearsal]” is wonderful in terms of both the artistic performance (singing and playing), and the resulting mix. But if I listen to the red book album version it just doesn’t sound as compelling. So, SQ aside (which a number of you have commented on) if the recording artist and ‘team’ can produce such a stunning live performance then why not achieve the same heights on the album?

Another example is comparing YouTube “HALOS - Full Performance (Live on KEXP)” with their album versions.

I’m not sure there is a scientific answer but I find it curious and wondered whether it’s just me.

 

[goat76]

My angle is slightly different though - because I’m comparing live performance with album version. I’ll give an example. The YouTube video titled “David Gilmour - Between Two Points (with Romany Gilmour) [Tour Rehearsal]” is wonderful in terms of both the artistic performance (singing and playing), and the resulting mix. But if I listen to the red book album version it just doesn’t sound as compelling. So, SQ aside (which a number of you have commented on) if the recording artist and ‘team’ can produce such a stunning live performance then why not achieve the same heights on the album?

It's not uncommon that the official release is overly processed in the production stage compared to a live in-studio session, that's why I love to watch videos where the bands play live for a radio show, "Tiny Desk Concerts", OneMic videos, "From the Basement", and similar stuff. But be aware that you also get the visual sensation which sometimes makes the sound appear to sound better than it actually is. Do you still find it better sounding when you shut your eyes?

 

[Palmer]

If you are streaming or posting audio on social media, a tool like meterplugs loudness penalty works wonders. My experience. A dear friend and streamer uses an Elgato Wave:3 and his audio always sound superb. Both tools are cheap and help a lot. Just throwing that out there, no affiliation.

 

[pablolie]

Aren't most YouTube videos 128k? Premium Spotify will murder that any time provided it's a good recording to begin with.
Mind you, I find many YouTube music videos totally enjoyable. Of course the visual adds to the experience even if the sound quality may not be the best of the best.

 

[Doodski]

Aren't most YouTube videos 128k? Premium Spotify will murder that any time provided it's a good recording to begin with.
Mind you, I find many YouTube music videos totally enjoyable. Of course the visual adds to the experience even if the sound quality may not be the best of the best.

Codecs av01.0.04M.08 (397) / opus (251)
I copied that from the YouTube stats for nerds right click option. Does this mean 397 and 251 for the bit rate?

 

[staticV3]

Codecs av01.0.04M.08 (397) / opus (251)
I copied that from the YouTube stats for nerds right click option. Does this mean 397 and 251 for the bit rate?
View attachment 396440

397 and 251 are internal ID codes denoting different formats.

The yt-dlp CLI will list all available formats for a given video, including bitrate and sample rate:


251 is by far YouTube's most-streamed audio format, which is 128Kbps Opus at 48kHz.

 

[Doodski]

No.
397 and 251 are internal ID codes denoting different formats.

The yt-dlp CLI will list all available formats for a given program, including bitrate and sample rate:
View attachment 396441

251 is by far YouTube's most-streamed audio format, which is 128Kbps Opus at 48kHz.

IC... Thank you.

 

[pablolie]

397 and 251 are internal ID codes denoting different formats.

The yt-dlp CLI will list all available formats for a given video, including bitrate and sample rate:
View attachment 396441

251 is by far YouTube's most-streamed audio format, which is 128Kbps Opus at 48kHz.

AV1 is the video encoding? And Opus seems a good 128k audio encoder?

 

[staticV3]

AV1 is the video encoding?

Correct.

And Opus seems a good 128k audio encoder?

Here's a Null test comparison between YouTube, Spotify, and MP3/AAC encoding:

Spectra:


Error spectrum:


I had to record Spotify via the Windows audio pipeline, instead of being able to download the audio straight from their servers like I can with YouTube.
I therefore tested YouTube with both direct download and playing in Chrome via Windows, to make sure that that would not skew the results.

As you can see, recording via Windows had no effect on audio fidelity.

Lastly, here's the PK Metric for each format:

 

[danadam]

251 is by far YouTube's most-streamed audio format, which is 128Kbps Opus at 48kHz.

Up to 160 kbps actually, e.g. Laurence Juber / Guitar Noir / In Your Arms:

# yt-dlp
251 webm  audio only      2 │   4.26MiB  164k https │ audio only          opus       164k 48k medium, webm_dash
# opusinfo
Average bitrate: 162.6 kb/s, w/o overhead: 161.5 kb/s

or Michael Murray / The Organ at St. Andreas Kirche / J.S. Bach: Prelude and Fugue in G Major, BWV 541:

# yt-dlp
251 webm  audio only      2 │  10.51MiB 162k https │ audio only         opus       162k 48k medium, webm_dash
# opusinfo
Average bitrate: 160.5 kb/s, w/o overhead: 159.4 kb/s

(though even at 128 kbps I don't think it would be "murdered" by anything )