You are here

Mastering again.

For everything after the recording stage: hardware/software and how you use it.

Re: Mastering again.

Postby Tim Gillett » Wed Jun 24, 2020 9:43 pm

Hugh Robjohns wrote:... I never judge sound quality from YT clips as there are too many unknowns...

Youtube now provides information as to whether it has reduced gain (loudness normalised) on uploaded files. Right click on the video being watched and "Stats for nerds" appears. YT does not increase level on "quiet" uploads but does lower ones considered to have too high average loudness - for a less unpleasant listening experience when listening to a random series of uploads of (previously) varying average loudness. If loudness normalisation has been used, that is gain reduction, the words "content loudness" appear along with the amount in db that the file has been gain reduced.

Unsurprisingly, the TIOBE file which I referenced has not been loudness normalised by YT, for its average level is very low, much lower than we would expect. I dont know what is its highest peak because I cant easily check it ( I would need to rerecord the entire movie in real time) but there's no reason to believe the upload is lower in average level than intended by the production company. So ... that an uploaded file from a major distributor is any lower in gain than was uploaded and intended seems highly unlikely.

When you said elsewhere the upload is "stupidly low' (in average level) it is indeed, but it's not the fault of YT who dont increase the average level of any uploads. It's "stupidly low" because most likely that was the file uploaded to YT!

And as we know, given that audio data compression as used on YT first throws away very quiet high frequencies, uploading to YT a file with much reduced high frequencies is the very last thing a good engineer would do. Again skilled mastering for the lower resolution destination format.

Unsurprisingly, compared to a loudness normalised DVD the audio on the DVD of the same remaster (2016) is also stupidly low .

But as already mentioned more than once, it's more than that. Within the file itself, the balance of speech frequencies is cockeyed. The vowel sounds are far too loud compared to the consonants. In technical terms, the frequencies below roughly 2 kHz are too high compared to those between 2kHz and 4 kHz, the latter carrying the important consonant sounds. It sounds muffled and indistinct. Turning the volume up helps somewhat but then the vowel sounds can become unpleasantly loud for many listeners.

It's interesting to compare the production company's promo sample to another older transfer YT upload. YT has loudness normalised the older transfer (second file) down by 8.6 db (see "stats for nerds") but it's still subjectively louder - and speech much clearer - than the official sample.The reason is that - again - the speech frequencies in the older transfer are much better balanced. More like normal speech. It doesnt have an excess of lower frequencies. YT does not change spectral balance.

And again, in the older transfer, the frequencies in the critical speech intelligibility band are relatively higher, so they are less likely to be reduced by the YT audio codec...compared to the official latest upload which risks those same highs, already too low, being reduced even lower by the YT codec.

https://youtu.be/9GsrDe96HUE Official sample 2016 remaster

https://youtu.be/IPLBAYvZVY8 Older transfer, same section at 1:05:19
Tim Gillett
Frequent Poster (Level2)
Posts: 2162
Joined: Wed Jan 30, 2013 1:00 am
Location: Perth, Western Australia

Re: Mastering again.

Postby Hugh Robjohns » Wed Jun 24, 2020 10:31 pm

Tim Gillett wrote:
Hugh Robjohns wrote:... I never judge sound quality from YT clips as there are too many unknowns...

Youtube now provides information as to whether it has reduced gain (loudness normalised) on uploaded files.

... So that aspect wouldn't be one of the unknowns to which I referred then, would it? ;)

One potential reason for a file being 'stupidly quiet' (as I described it) could be that it came directly from a broadcast master compliant with the R128 (and equivalent) specs, so that would make it 9 or 10dB lower than the current target loudness for YT. In this case, though, it sounds even lower than that to me... But, like you, I'm not going to download the whole film just to check!

I might be disappointed if whoever authorised uploading the file didn't take that into account, but not surprised -- there's nothing that forces uploads to be performed by technically competent staff.

And these are just some of the real unknowns. What was the real source of the material? Who undertook the upload and did they know what they were doing?

Of much greater significance to a sound quality problem like the specific one identified in this thread, is whether the audio source was the master linear audio file, or was it already in a data-reduced format like AC3 (Dolby Digital) -- perhaps because the video was ripped from a commercial DVD etc.

If that was the case we have unpredicatable codec concatenation artefacts to consider which could very easily account for some of the loss of intelligibility. And that, combined with the effect the low level would have on YT's own lossy codec could account for more damage.

So, all in all, proclaiming on the apparent sound quality of source material based purely on what is heard from a YT clip is, in my humble opinion, somewhat foolhardy to put it politely! There really are just too many unknowns....

Offsetting that in this case is that it appears that some customer reviews of the commercial disc also indicate some intelligibility problems, so clearly something is amiss, but to know exactly what would really require an examination of the whole production process signal path.

As a guess, I suspect the production company insisted on a hiss- and crackle-free audio track to go with the squeaky-clean images, and threw the baby out with the bath water in the process -- perhaps (hopefully) going against the advice of their technicians.

...but we'll never know for sure.
User avatar
Hugh Robjohns
Moderator
Posts: 27858
Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Technical Editor, Sound On Sound

Re: Mastering again.

Postby Tim Gillett » Wed Jun 24, 2020 11:47 pm

Hugh Robjohns wrote:
...As a guess, I suspect the production company insisted on a hiss- and crackle-free audio audio track to go with the squeaky-clean images, and threw the baby out with the bath water in the process -- perhaps (hopefully) going against the advice of their technicians.


Now why didnt I think of that? ;)

Hugh Robjohns wrote:but we'll never know for sure.

Yes, that we can never know for certain (about this matter) is a certainty...

Now there's a conundrum to ponder... ;)
Tim Gillett
Frequent Poster (Level2)
Posts: 2162
Joined: Wed Jan 30, 2013 1:00 am
Location: Perth, Western Australia

Re: Mastering again.

Postby Urthlupe » Thu Jun 25, 2020 3:04 pm

Wow...... more circular impenetrable guff, and as ever, entertaining in a car crash kinda way (which explains how I got this far down the thread, sorta).

Reading it is for me like one of those things you do then dislike yourself for afterwards.

:mrgreen:

Loopy
User avatar
Urthlupe
Frequent Poster
Posts: 518
Joined: Fri Sep 20, 2002 12:00 am
Location: West Midlands/Pembrokeshire UK

Re: Mastering again.

Postby Tim Gillett » Fri Jun 26, 2020 9:43 am

Martin Walker wrote:
Tim Gillett wrote:It sometimes gets worse. A perfectly fine, old movie soundtrack - which is 99% speech driven - gets the "remastering" or "cleanup" treatment and this is the result:

https://youtu.be/9GsrDe96HUE

Ye gods - that IS atrocious! :crazy:


Martin

Thanks Martin, yes a shame especially because many interested in the old film are probably elderly themselves with some understandable hearing loss, making it even harder for them to understand the already muffled audio.

I'm especially tuned into this sort of thing. I worked as a tech with the hearing impaired and then later in talking books for the blind. Obviously in both areas speech intelligibility is really important.
Tim Gillett
Frequent Poster (Level2)
Posts: 2162
Joined: Wed Jan 30, 2013 1:00 am
Location: Perth, Western Australia

Previous