I

Ok I finally checked it. It just sounds like you didn't maximize your track for -14LUFS-I -1db and the track got quieter and a bit less detailed after posting to YT. Nice fado piece btw. What is the site videy's upload guideline suggestions for an audio track?

 

@Crowyote Yeah I noticed that too, I turned up the volume for youtube in my comparison but did not realise it also affected audio quality. I guess at the low bitrate that is true.

I used YT downloader to get the audio values from YT, there is also stats for nerds that you see John Dennis using in a screen shot above. Btw I had a look at vimeo, for a random 1080p video I chose the video and audio bitrates are twice that of youtube. around 5000kbit/s for video, and 250kbit/s AAC. You might want to try it, but Vimeo not so great for discoverability. Depends on your needs

I have a friend who graduated from film school and competes in festivals. She only uses Vimeo, and I guess that it why. I don't really because they charge for space, but I'm tempted to start an account or two for certain media I'd like to display.

Ok I finally checked it. It just sounds like you didn't maximize your track for -14LUFS-I -1db and the track got quieter and a bit less detailed after posting to YT.

Very astute observation! You are well on your way to mastering Youtube (such as it is...); in all honesty I haven't run up/down audio tests on Youtube for some time. But their Opus sounds quite good to me. I hope you will continue to share your results; they will be a great resource.

I will say that I like the ability to have a lower peak like -16 or -18 even tho I've gotten used to -14 there is a loss of headroom with that level,

I think you can upload -6dBTP at -15LUFS to Youtube without incurring a loudness penalty. That frees up some dynamic range (LRA), which can be leveraged with just a little compression and a transparent limiter, giving the acoustic richness I love to hear.

To that purpose, I would like to introduce you to my very favorite, always-on VST plugin, the Barricade 4 by Toneboosters. With mathematically flawless displays and meters, intuitive controls, and a couple of amazing presets, I never start a project without it. The great news is, Version 3 Toneboosters is now free (!) so you can check out its essential functions without plunking down $20. I know this sounds like a commercial; I was a beta tester for his very first EBU R128 meters, and I've been hooked on them ever since.

Here's a short demo; It was recorded in the orchestra pit, and unfortunately there is a little trash from the stage monitors, but it demonstrates Barricade's faithfulness with acoustic instruments.

Thanks so much! If there's one thing I like it's a good transparent VST, if there's one thing I love it's a FREE transparent VST! :D

I'll consider those limits for more complex material or maybe for pre-"mastering" for an upcoming vinyl release. I'm kinda sticking to this formula since it works for now. I still haven't tried the "loudness logs" on Vegas . . .but it still seems so much more efficient to strip the audio from Vegas "master" in Reaper and reinsert it since Izotope is pretty CPU intensive and so are all the Vegas plug-ins.

Also raise your hand if you also can stand the automatic plugins that appear when one starts an audio track? I can't tell you how long it took at first to figure out why tracks sounded different and worse after exporting from VP until I figured it out. Anyway to disable (compressor, noise gate, etc)?

You can delete the plugins from the tracks and then save it as a new track default. They shouldn't be affecting the sound of the track, though.

1. I misread your post because you said "I will say that I like the ability to have a lower peak like -16 or -18 even tho I've gotten used to -14" when you meant to say "integrated" LU. I'm sure it was just a typo.

But indeed, -12dB RMS with -6dBFS peaks are pretty normal for acoustic (non-streaming) music recordings. Jacking acoustic peaks to -1 dB sounds harsh and is hard for my ears to take, even with the best mics. Also, the re-encoding of good audio to AAC often turns edgy peaks to clipped distorted bursts, so I avoid that particular gimmick like the plague, even if it makes my stuff less "competitive" in the internet jungle. For my ears, battered by fifty years of recording experience, I don't go there; yes that is only my preference. Do you agree, @rraud?

2. As far as the audio plugin chain in its default state, I ran empirical bitwise tests years ago (pre-Magix), revealing no measurable waveform differences; they were identical using uncompressed source. If you are hearing something different, perhaps you could construct your own quantified tests and share the results. It would be enlightening.

3. Personal preferences being what they are, the last documentation I have of the loudness standards and penalties from various streaming services is here. I have found automated social loudness bots to be a bit inconsistent in their application.

https://youlean.co/loudness-standards-full-comparison-table/

battered by fifty years of recording experience,

About the same here @Musicvid , I stated playing music in bands when I was 14 and moved on to FOH sound and recording. I was involved in other high SPL activities when I was young. My ears are focked up as well. I have adapted though and can still mix on a pro level.

Rock music material streaming at -14 LUFS (integrated) loudness, (the recommended streaming loudness level) will typically have occasional 'true peaks' approaching - 2.0 dBFS.
Many folks confuse decibel (dBFS peak) meters for loudness LU/LK meters, where 'Integrated' and 'True Peak' are the important factors. Broadcast submission loudness levels for narratives and such must be significant lower. -23/24 LUFS (integrated.. which means the entire program).

I usually use the Sound Forge Pro statistics tool to measure the integrated LU and true peak factors, which is much faster than real time'. Unfortunately, Vegas cannot render an audio statistics log. But there are plug-ins that can set the overall level automatically. They are not perfect though and many calculate the loudness based on only a few seconds of playback. I recommend double-checking with a trusted meter before submitting anything, Most music streaming services will re-render if the levels are off-spec. Some will boost or reduce the levels and some will only reduce the level if it's over -14 LU.

Rock music material streaming at -14 LUFS (integrated) loudness, (the recommended streaming loudness level) will typically have occasional 'true peaks' approaching - 2.0 dBFS.
Many folks confuse decibel (dBFS peak) meters for loudness LU/LK meters, where 'Integrated' and 'True Peak' are the important factors. Broadcast submission loudness levels for narratives and such must be significant lower. -23/24 LUFS (integrated.. which means the entire program).

Couldn't agree more; of course with a century of combined audio engineering experience between us, I have come to expect that.

Interesting that the newer EBU R128 S1 European TV standard also sets -18 LUFS as a Short Term loudness target (LS Max).

For those interested in details of loudness units, LKFS (US) is slightly lower than LUFS (EU), so that anything passing ATSC A/85 will also consistently meet EBU R128 without checking.

Of course, as far as the FCC is concerned, it's all a farce. Major broadcasters still push commercial levels well beyond the NAB best practices (and the law) and pretend we won't notice. I do notice; I first started tracking this stuff in 2011 when I developed hyperacusis. Fox is still one of the worst offenders, as it was then.

1. I misread your post because you said "I will say that I like the ability to have a lower peak like -16 or -18 even tho I've gotten used to -14" when you meant to say "integrated" LU. I'm sure it was just a typo.

But indeed, -12dB RMS with -6dBFS peaks are pretty normal for acoustic (non-streaming) music recordings. Jacking acoustic peaks to -1 dB sounds harsh and is hard for my ears to take, even with the best mics. Also, the re-encoding of good audio to AAC often turns edgy peaks to clipped distorted bursts, so I avoid that particular gimmick like the plague, even if it makes my stuff less "competitive" in the internet jungle. For my ears, battered by fifty years of recording experience, I don't go there; yes that is only my preference. Do you agree, @rraud?

2. As far as the audio plugin chain in its default state, I ran empirical bitwise tests years ago (pre-Magix), revealing no measurable waveform differences; they were identical using uncompressed source. If you are hearing something different, perhaps you could construct your own quantified tests and share the results. It would be enlightening.

3. Personal preferences being what they are, the last documentation I have of the loudness standards and penalties from various streaming services is here. I have found automated social loudness bots to be a bit inconsistent in their application.

https://youlean.co/loudness-standards-full-comparison-table/

Thanks!!!

Yes, I made a typo. I'm an artist and an abstract one at that, not an engineer by any standards, so please give me some leeway ;p.

I really appreciate your input and discussion on this thread. I wonder what you think of the audio quality of Soundcloud's player: https://soundcloud.com/perrybrooksnicholsnylon/ramkali-romance-open-d-major

I've been moving away from Soundcloud as they have started to run way too many ads and the storage limits are frustrating (on my 4th profile!), but for a time I built a considerable following there with my first profile. Bandcamp has no storage limit, but it was only until we had this discussion that I realized how poorly it's representing the quality of my music via the streaming player.

As far as using Reaper to generate -14LUFS-I and -1 db Tp, it seems accurate and with the dry run option providing analytics on the duration of a rendered track. Before they came out with dry run renders, I was rendering complete tracks over and over until I hit the target and it was wasting tons of hd space. I agree with you about the harshness of the above limits which is why I want to discuss and investigate the range of output settings that work.

For now, since most of my tracks are acoustic I've learned that using Izotope Ozone 9's Tonal Balance control can really reduce the harsh peaks and avoiding compression altogether - when it doesn't contribute much to the sound - helps a bit. I tend to like to have a little bit of stereo independence on the transients which can add harshness at the maximizer stage, I suppose. I probably overuse the imager module as I go about 20ms stereoize with "anti-phase" engaged, and tweak all the bands until I get about the same peak on both channels targeting -1dbTP, (but not in the same places if it's a successfully recorded track!) - this is about the most concentration intensive part of my process and spreading the high band to the max in order to create more spatial illusion definitely risks adding a sharper tone to the mix.

I know very little about which video formats use what and especiallhy those that are using AAC, but I guess that 's mostly mp4 - with FB or IG uploads there's no other choice tho. I want to look into opus and XAVC, but I'm just using AVI for now since it seems to work.

I'll adjust the meters and try those other limits you've suggested as I get closer to finishing my album.

battered by fifty years of recording experience,

About the same here @Musicvid , I stated playing music in bands when I was 14 and moved on to FOH sound and recording. I was involved in other high SPL activities when I was young. My ears are focked up as well. I have adapted though and can still mix on a pro level.

Rock music material streaming at -14 LUFS (integrated) loudness, (the recommended streaming loudness level) will typically have occasional 'true peaks' approaching - 2.0 dBFS.
Many folks confuse decibel (dBFS peak) meters for loudness LU/LK meters, where 'Integrated' and 'True Peak' are the important factors. Broadcast submission loudness levels for narratives and such must be significant lower. -23/24 LUFS (integrated.. which means the entire program).

I usually use the Sound Forge Pro statistics tool to measure the integrated LU and true peak factors, which is much faster than real time'. Unfortunately, Vegas cannot render an audio statistics log. But there are plug-ins that can set the overall level automatically. They are not perfect though and many calculate the loudness based on only a few seconds of playback. I recommend double-checking with a trusted meter before submitting anything, Most music streaming services will re-render if the levels are off-spec. Some will boost or reduce the levels and some will only reduce the level if it's over -14 LU.

Thanks for your input! I agree about Broadcast levels, television is all over the place. Fox is an offender on so many levels, but why is HBO's Bill Maher so quiet compared to the rest of the broadcast.

I've not tried Sound Forge Pro at all, but if I can get a decent VP20 upgrade offer one of these days it should come with.

What I am trying to share is that one needs to not push the published loudness limits too far. Perfect example: your Opus peaks as encoded by Youtube are now louder than -1dBTP because their re-encoding produced intersample peaks in the form of quantization noise, a phenomenon I have dealt with for many years in my own work. At 73, I can hear slight distortion in your recording. Just back away a bit.

Check the other guy (eduardo's) guitar work with loudness meters. Acoustic guitar, nor any other stringed acoustic instrument, not even a concert harp or grand piano, needs or necessarily sounds best at such high peak levels. A good LRA is "much" more important with acoustic instruments. His TP and Integrated levels are much lower than Youtube's limits, and yet it sounds so good to me I've been playing it over and over.

There "is" a place for loudness competition on Youtube. However, it is not with acoustic or even orchestral performances. Before loudness standards, it was all around -12dB RMS, -6dBFS peaks on recordings. I hope this doesn't sound insensitive to your good work; I'm currently finishing my third (and last) brew this evening. Best.

 

I really appreciate your input and discussion on this thread. I wonder what you think of the audio quality of Soundcloud's player: https://soundcloud.com/perrybrooksnicholsnylon/ramkali-romance-open-d-major

That's also a 128kbit/s mp3 like bandcamp. Not sure why they're all using mp3's at such a low bitrate and not a better codec, maybe licensing, with mp3 possibly having no license cost. There was a popular piece of music that was a great demonstration of the limitations of 128kbit/s

It may have been 1 of these 2 songs

With source being a CD, the 128kbit/s mp3 encode falls apart, a cacophony of distortion , at 192kbit/s it sounds much better, and 320kbit/s better again. I'm not sure if the mp3 codec see's your music as overly complex or not. It certainly doesn't 'fall apart' like some music does at 128kbit/s

I've been using Soundforge (don't laugh) v7 for years, mostly for the Statistics function. I run it on Vegas mixdowns and don't use it to edit anything, just to determine what adjustments I need to make in the Vegas mix. 1st I look at the channel peaks and generally aim for a max of -0.5 db. Then look at the rms values for each channel aiming for the same ballpark as other pieces out there in the same genre. Rock with screaming vocals will come in typically at around -10 to -12 db rms power which is pretty loud. For most of the Jazz piano stuff I shoot, I aim for -19 rms +/- 2 depending on if it's intended to be delicate or strident, tempo, etc. Here's what the Sony Red Car soundtrack which leans towards orchestral looks like in v7:

Notice that there's a min and max peak for both channels to eyball with the overall peak being -2.2. Also notice there's a dc offset present which can cause overloads with filtering and eq... I avoid that like the plague.

Just downloaded an SF 16 trial and this screen has gotten much better... it's got all the same info but continues on with a summary of actual peak (highest of min and max) for each channel and also adds a line for filtered rms without dc offset. It's also got the noise/pop/click reduction pack thrown in which is really quite excellent... the old Sony one I have cost almost as much as Sound Forge back then.

For the catv market in the us that I also do stuff for, the requirement is that maximum peaks not exceed -6 db. So when I finish, lets say a song clip for YouTube as I've described above, I render another version for tv by throwing an audio FX onto the Vegas master slider (which happens to also exist in SF)... Wave Hammer Surround. Here's what the preset I use looks like:

They could of just as well called it Wave Crusher.

Here's my problem for tomorrow... gotta shoot my wife playing this on an acoustic guitar. I just pulled Joni's performance off of YouTube to get an idea of what loudness to aim for... around -17 db rms power:

update: shot, mixed, graded, and uploaded to YT... ended up aiming for -18 db rms power as the vocal has a wide and powerful low-end dynamic compared to the original.

What I am trying to share is that one needs to not push the published loudness limits too far. Perfect example: your Opus peaks as encoded by Youtube are now louder than -1dBTP because their re-encoding produced intersample peaks in the form of quantization noise, a phenomenon I have dealt with for many years in my own work. At 73, I can hear slight distortion in your recording. Just back away a bit.

Check the other guy (eduardo's) guitar work with loudness meters. Acoustic guitar, nor any other stringed acoustic instrument, not even a concert harp or grand piano, needs or necessarily sounds best at such high peak levels. A good LRA is "much" more important with acoustic instruments. His TP and Integrated levels are much lower than Youtube's limits, and yet it sounds so good to me I've been playing it over and over.

There "is" a place for loudness competition on Youtube. However, it is not with acoustic or even orchestral performances. Before loudness standards, it was all around -12dB RMS, -6dBFS peaks on recordings. I hope this doesn't sound insensitive to your good work; I'm currently finishing my third (and last) brew this evening. Best.

Interesting info, and thanks for your honesty. I hear the distortion too, but I did not think it too bad compared to the original. Of course, I'm probably overusing Izotope Ozone tbh. I'll try for the lower target suggested earlier as I'm not trying to win loudness wars, just trying to avoid their maximizer changing the dynamics of the track . . . but yeah, I need to drop that peak from -1dbTP, I sense.

 

I really appreciate your input and discussion on this thread. I wonder what you think of the audio quality of Soundcloud's player: https://soundcloud.com/perrybrooksnicholsnylon/ramkali-romance-open-d-major

That's also a 128kbit/s mp3 like bandcamp. Not sure why they're all using mp3's at such a low bitrate and not a better codec, maybe licensing, with mp3 possibly having no license cost. There was a popular piece of music that was a great demonstration of the limitations of 128kbit/s

It may have been 1 of these 2 songs

With source being a CD, the 128kbit/s mp3 encode falls apart, a cacophony of distortion , at 192kbit/s it sounds much better, and 320kbit/s better again. I'm not sure if the mp3 codec see's your music as overly complex or not. It certainly doesn't 'fall apart' like some music does at 128kbit/s

 

I think I figured out that 128kbps is the standard because of Facebook. They've limited all of their streaming audio and videos to 16bit/128kbps AAC. If you upload one of greater quality it will have artifacts and distortion. I experience this all the time because instagram allows videos of at least 16/256kbps and sometimes it crossposts a video to my FB page. Everytime I upload a video with any quality higher than 128kbps and it crossposts, aliasing noise etc enters the equation. We're all trapped in Zuck's Tesseract, I suppose.

Here's my problem for tomorrow... gotta shoot someone playing this on an acoustic guitar. I just pulled Joni's performance off of YouTube to get an idea of what loudness to aim for... around -17 db rms power:

Interesting, I'll see if Reaper has a setting that indicates DC offset, because that seems like the most interesting thing to learn about here . . .

I've not tried Sound Forge Pro at all, but if I can get a decent VP20 upgrade offer one of these days it should come with.

@Crowyote, none of the current versions of Vegas include Sound Forge Pro. Vegas Pro and Vegas Post include the Audio Studio version of Sound Forge. However AS does not have the 'Statistics' tool. Otherwise it is good application to have.

rraud, Magix forums moderator

What I am trying to share is that one needs to not push the published loudness limits too far. Perfect example: your Opus peaks as encoded by Youtube are now louder than -1dBTP because their re-encoding produced intersample peaks in the form of quantization noise, a phenomenon I have dealt with for many years in my own work. At 73, I can hear slight distortion in your recording. Just back away a bit.

Check the other guy (eduardo's) guitar work with loudness meters. Acoustic guitar, nor any other stringed acoustic instrument, not even a concert harp or grand piano, needs or necessarily sounds best at such high peak levels. A good LRA is "much" more important with acoustic instruments. His TP and Integrated levels are much lower than Youtube's limits, and yet it sounds so good to me I've been playing it over and over.

There "is" a place for loudness competition on Youtube. However, it is not with acoustic or even orchestral performances. Before loudness standards, it was all around -12dB RMS, -6dBFS peaks on recordings. I hope this doesn't sound insensitive to your good work; I'm currently finishing my third (and last) brew this evening. Best.

Thanks again for the insights. Apparently, I didn't have "True Peak" selected in my dry run statistics this whole time, so I was likely going over -1dBTP all this time. A few runs verified that. After it would do the run set on TP the final reading would jump up a bit.

What do you consider a good LRA?

Sadly I can't find the free version of that plugin you mentioned.

1. Type "Toneboosters Free" into a Google Search.
2. Click on the first link that comes up.
3. 
4. Scroll down the page to the link to the Software Updates Page. Click it.
5. Scroll down to the bottom of the page for the Toneboosters 3 download package.

At least that's how I found it...

LRA stands for Loudness Range, also called Dynamic Range. For acoustic music, wouldn't you like it to be as great as possible without pushing one of the predetermined limits, such as True Peak or ILU? I think I would. @edoardo-l put up a really amazing 9dB LRA with rather conservative TP and ILU scores.

Really Crowyote, all this stuff does is to set up a space to keep your loudness parameters within. You are not competing with overcompressed electronic guitar jams on Youtube, at least I hope not! If by some chance you are searching for a holy grail setting by micromanaging your meters, it means it's time to stop looking and listen.

Best of success with your project.

1. Type "Toneboosters Free" into a Google Search.
2. Click on the first link that comes up.
3. 
4. Scroll down the page to the link to the Software Updates Page. Click it.
5. Scroll down to the bottom of the page for the Toneboosters 3 download package.

At least that's how I found it...

LRA stands for Loudness Range, also called Dynamic Range. For acoustic music, wouldn't you like it to be as great as possible without pushing one of the predetermined limits, such as True Peak or ILU? I think I would. @edoardo-l put up a really amazing 9dB LRA with rather conservative TP and ILU scores.

Really Crowyote, all this stuff does is to set up a space to keep your loudness parameters within. You are not competing with overcompressed electronic guitar jams on Youtube, at least I hope not! If by some chance you are searching for a holy grail setting by micromanaging your meters, it means it's time to stop looking and listen.

Best of success with your project.

Nah, I'm just trying to get my tracks more in the target range so that the masters sound more like they would on YT or some other service that might adjust them. Like you said, there are lower levels that won't be affected by the "loudness penalty," and it's a learning process now that I've gotten used to one thing to bring them back down and leave some headroom. My genre (American Primitive Guitar) is generally a little more aggressive than Eduardo's and he's recording musicians in really wonderful acoustic spaces, a luxury I don't have. The current track I'm working on is ranging from 4.8LRA to about 18! I'm trying to get it to -16LUFS but not much luck keeping -2dbTP, I'll keep trying tho.

I've been using Soundforge (don't laugh) v7 for years, mostly for the Statistics function. I run it on Vegas mixdowns and don't use it to edit anything, just to determine what adjustments I need to make in the Vegas mix. 1st I look at the channel peaks and generally aim for a max of -0.5 db. Then look at the rms values for each channel aiming for the same ballpark as other pieces out there in the same genre. Rock with screaming vocals will come in typically at around -10 to -12 db rms power which is pretty loud. For most of the Jazz piano stuff I shoot, I aim for -19 rms +/- 2 depending on if it's intended to be delicate or strident, tempo, etc. Here's what the Sony Red Car soundtrack which leans towards orchestral looks like in v7:

Notice that there's a min and max peak for both channels to eyball with the overall peak being -2.2. Also notice there's a dc offset present which can cause overloads with filtering and eq... I avoid that like the plague.

Just downloaded an SF 16 trial and this screen has gotten much better... it's got all the same info but continues on with a summary of actual peak (highest of min and max) for each channel and also adds a line for filtered rms without dc offset. It's also got the noise/pop/click reduction pack thrown in which is really quite excellent... the old Sony one I have cost almost as much as Sound Forge back then.

For the catv market in the us that I also do stuff for, the requirement is that maximum peaks not exceed -6 db. So when I finish, lets say a song clip for YouTube as I've described above, I render another version for tv by throwing an audio FX onto the Vegas master slider (which happens to also exist in SF)... Wave Hammer Surround. Here's what the preset I use looks like:

They could of just as well called it Wave Crusher.

I'm only seeing DC offset in Reaper when it comes to Dithering. I'm don't dither any of my material. Not really sure if it's necessary or what that's all about.

Use a limiter, such as already exists in Vegas (Wave Hammer), or the one I suggested (Toneboosters), and peaks are easily tamed without touching anything else.

Thanks for your questions, experience will answer them for you, and good luck with your Youtube venture! I honestly don't know much more than I've shared, so have a ball!

@Crowyote

Don't get too hung up on exact numbers. Definitely watch your peaks. The more you compress the audio the loudness range will be smaller. But that is highly dependent on the sound source. I agree with @Musicvid that a solo string instrument would sound terrible if you tried to match loudness of fully mastered EDM track. You would have to compress it to death.

You should use dither when going down in audio bit depth. Let's say you record in 24 bits and you want to render to 16 bits. The dither affects mainly the most quiet parts. Chances are you would here no difference when listening. It adds low level (very low) noise to mask the quantization errors.

That is exactly correct, @misohoza

+1

@Crowyote

Don't get too hung up on exact numbers.

@misohoza Use the numbers as a starting point but base you final decisions on listening. Rms numbers are not weighted for perceptual differences in loudness as frequency distribution varies. LUFS however, are weighted for frequency distribution. Still need to listen, however, since I don't think LUFS accurately treat tempo variations. But I'm probably going to upgrade to sf16 if only for the LUFS readout.

Definitely watch your peaks.

This was mentioned earlier in passing, but looking more closely at sf16, I see they give 5 peak statistics: min and max channel sample values, max channel true peak, max filtered channel true peak, and max true peak (overall at the bottom) . The channel true peak is the greater of min and max absolute sample value for each channel. The filtered channel true peak would be the peak after applying a high-pass filter, presumably 20hz, in the presence of a dc offset. Same for rms and filtered rms.

Btw, if a dc offset got past your preamps, you can throw a 20 hz hp filter on the track yourself in Vegas or SF. In which case the true and filtered measurements would come out almost the same and dc offset would be minimized. Here's how to do it with the default Vegas track eq...

I don't recommend going any steeper than 12 db/oct.

@Crowyote

Don't get too hung up on exact numbers. Definitely watch your peaks. The more you compress the audio the loudness range will be smaller. But that is highly dependent on the sound source. I agree with @Musicvid that a solo string instrument would sound terrible if you tried to match loudness of fully mastered EDM track. You would have to compress it to death.

You should use dither when going down in audio bit depth. Let's say you record in 24 bits and you want to render to 16 bits. The dither affects mainly the most quiet parts. Chances are you would here no difference when listening. It adds low level (very low) noise to mask the quantization errors.

Not true at all. You don't know my process at all. I use very little compression, and I'm a very dynamic player who often plays pretty close to the mic just below the level of clipping. Sometimes I use one in the loops of my stereo mic preamp, a very low mix Effectrode LA-1A pedal and/or another tube compressor with a light mix. I try to avoid lots of compression in mixing. I do use a considerable amount of imaging since I don't have an incredible space to record in at present.

I rendered my latest track to 16 LUFS -1.2dbTP, and I was only able to get it down by adding a slight slope of compression to the bass and some tone contour on the highs (the guitar has an excessively bright almost harsh toned maple body) . It was getting to much higher levels like 15 LUFS -1 etc.
https://crowyote.bandcamp.com/track/moonflower-dance-yaman-kalyan

Of course you might not be able to tell with the mp3 player there. I should have the video ready in a week or so, so we'll have to see how it pans out when rendered. I'm hoping I can afford those harsh bursts and peaks that MUSICVID was referring to when it finishes up.

I'm a very dynamic player who often plays pretty close to the mic just below the level of clipping.

Might try pulling your mic back a bit and if it is cardioid, not pointing it directly at the hole. Sometimes its best to let guitars to build a little acoustic space. I might add a little paragraphic shelf around E2 (~80hZ) to tame that open string. That's about all I have to offer. Entirely your preference, of course!