Highest possible audio quality for YouTube

[Entheo]

Full disclosure - I haven’t installed Shotcut yet because I have one main question before committing to installing and learning the program. I currently make a lot of musical videos (piano and drumming videos) and upload them to my respective YouTube channels. For example, I record my drumming with a Zoom Q2n, mix the drum track with the drumless track in Audacity, save as a WAV file, and do final production (with HD video) in the old Windows MovieMaker (Shotcut would replace WMM). I then export as MP4 and upload into YouTube.

My query is this: once uploaded into YouTube, the sound quality is considerably diminished from the original WAV audio mixdown. I’m not sure if this is due to WMM limitations or file compression in YouTube, but I’m hoping there are audio tools or formats in Shotcut that would provide me with less degradation in audio quality upon uploading to YouTube.

Any help or advice is appreciated - thanks in advance.

[Austin]

Shotcut can export to pretty much any format. This is one of its strengths… pretty much anything that FFmpeg can encode, Shotcut can encode. And it provides you with extensive manual controls if you need it.

The audio compression by Windows MovieMaker is probably the biggest choke point in your current workflow.

With Shotcut, you could export as a Matroska file with H.264 video and WAV audio, meaning your mixdown master would go to YouTube with zero loss. Can’t beat that. Any loss from there would be YouTube’s re-compression. (WAV audio can be selected on the “Audio” tab by choosing the pcm_s16le or pcm_s24le codec, depending on whether your sources are 16- or 24-bit.)

There is the FLAC option if you want lossless files that are around half the size of WAV.

If you need an even smaller file to upload, you can send Opus audio to YouTube, which is about as good as it gets for lossy formats:

[Entheo]

wow, thank you so much, Austin! looks like i’ll be moving forward with Shotcut!

[Entheo]

Hi again Austin - quick comment and a question. I ran a test with your recommended audio settings and uploaded to YouTube. I couldn’t really hear any audio quality difference between the WMM version and the Shotcut version, so I’m wondering if the problem is more on the YouTube compression side of things. But my next video will use Shotcut for sure!

Question - for the audio codec, should I choose “libopus” (in your screen shot) or “pcm_s16le”? My original audio file is in WAV format.

Thanks again!

[Austin]

Hmm, I was hoping that WMM compression was the biggest problem. That’s pretty alarming that YouTube compression is messing up your sound that much. On the flip side, if your video is 1080p or lower resolution, then you’re getting one of YouTube’s lower bitrates along with AAC audio, and AAC is not exactly a strong performer on drums, particularly high-frequency stuff like cymbals. AAC can get shrill and wobbly.

To clarify, pcm_s16le means 16-bit WAV output, which would basically mean lossless quality. If you choose libopus, the audio will be compressed and lose some detail, but it will be the least amount of loss among the lossy formats. If you have disk space and Internet bandwidth to spare, then pcm_s16le will produce the best sound at the expense of larger file sizes. There is also FLAC or ALAC for compressed lossless. If the file sizes need to be even more conservative, then libopus is the ticket.

There might be one more option… If your videos are 1080p, a common trick is to export them as 1440p to make YouTube give you a higher bitrate or a better codec, which includes audio. Try making another 1080p video the usual way again, but then do this in addition to the previous settings when it’s time to export and see if it sounds better when uploaded to YouTube:

[Entheo]

thank you, Austin - i will give this a try and report back!

by way of background: my Zoom Q2n is set at Video 1080·30, Resolution 1920×1080, frames/second 30 and audio is 48k 24 bit. the playalong tracks are usually mp3 and come in at 44.1k, but i export the mixdown as WAV.

also, if you’re interested (headphones rec’d) -
here’s a WAV file uploaded to my SoundCloud channel: Stream Get It On (big band) by Ed Kriege, Drums and Piano | Listen online for free on SoundCloud
here’s the YouTube upload: https://youtu.be/9m8jHEWMlJ8
pretty clear there is loss of audio quality.

[Entheo]

hi Austin - reporting back. I created a version with your video and audio settings and uploaded it to YouTube. There was possibly a slight improvement in audio quality, but still nowhere near the quality of the WAV file. I should also mention that, when I upload a video to Facebook, the audio quality is on par with the WAV file (after a period of time FB compresses the video to a very low quality, like 720).

So, I think I have to conclude that perhaps there’s no way of getting around YouTube’s compression algorithms.

Thanks so much for your help - I learned a lot!

[Austin]

Bummer, that’s unfortunate to put a lot of work into an art that YouTube crushes. Just to verify, if you pull up right-click > Stats for Nerds in the YouTube player, does it show VP9 or AV1 for the video codec, and Opus for the audio codec? If it doesn’t, then your video still isn’t getting that next-level quality bump. Maybe your player didn’t pull the 1440p version by default, or it’s still being encoded in YouTube’s queue if that’s the case.

If all that checks out and YouTube truly is destroying your audio, then at least you know you’ve done everything you can do and the problem isn’t on your end. Cool experiment, thanks for sharing the results!

[Entheo]

Hi Austin - here’s what comes up:

Codecs vp09.00.51.08.01.01.01.01.00 (248) / opus (251)

does that look correct?

ps - I selected Quality = 1440p

[Austin]

Yup, you got the good stuff. If it still sounds bad, that’s on YouTube and there’s nothing more you can do about it. I’m surprised that Opus would produce distortion that’s super-noticeable, even from YouTube. I don’t know the link to your latest Opus upload, but is it like blatantly bad sounding? Or do I have to put on headphones and dive into squint-mode kick-drum-solo-land to hear the difference? Also, which part sounds bad? Your drums that were recorded in WAV, or the backing track that was already a crushed MP3 from the start? I wouldn’t have high expectations for that MP3 holding up through YouTube…

[Entheo]

although I’ll be deleting the test recording soon, here’s an unlisted link to it:

- YouTube

It’s not that it sounds terrible (e.g., distorted), it’s that its quality is considerably less than the WAV file (see the Soundcloud link I posted earlier for comparison, and yes - i think headphones make the differences pretty obvious) and it’s considerably less than the MP4 versions posted to Facebook (YT versions are a bit dull, quieter, and not as sharp and full as the other versions). It’s true that I’m usually mixing my WAV drum recording with an mp3 backing track, but my final mix (saved as WAV) is quite satisfactory.

Don’t we have to conclude that, in spite of a good WAV audio and the highest quality codecs being used, that YT is the culprit in its compression?

Thanks again!

[Austin]

Oooooooooooooookay, I finally understand what’s happening now. Sorry, this thought didn’t cross my mind earlier.

Groovy stuff by the way! I dig it. Great results from a Zoom, too.

Check out the Stats for Nerds for your video:

Notice this line:

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

YouTube has a LUFS normalization system in place. Remember the days when you’d be watching a TV show, then a commercial would come on that was 120 dB louder than the show and it would drive you insane? YouTube drops the hammer on that nonsense by turning down any video whose “average volume” is louder than -14 LUFS. (TV stations today have also dropped the hammer by requiring all material be normalized to -23 or -24 LUFS, depending on whether you’re in USA or EU.)

The Stats for Nerds box tells us that your audio is 4.8 dB above the -14 LUFS threshold. So, YouTube turns your audio down during playback by 4.8 dB so it matches all the other videos out there. Soundcloud does not clamp you. So you are indeed hearing a quieter version of your track on YouTube, which psycho-acoustically gets interpreted as less “bright, full, energetic”. But it’s only a volume change.

I did a soundcard loopback rip of your audio from both YouTube and Soundcloud, then brought both tracks into Shotcut (just to be fast). I put a Gain filter on the Soundcloud version that dropped its volume -5 dB. (The relationship between RMS dB and LUFS dB is not always 1:1.) Now that YouTube and Soundcloud are level-matched in Shotcut, they pretty much sound the same if I do an A/B test between them. There is ever-so-slight lossy compression apparent on the YouTube version, probably because Soundcloud offered a higher bitrate. However, I’m blasting this on Dynaudio BM5A nearfields and I’m specifically looking for problems. Nobody else will notice a difference between these tracks once they’re level-matched.

So, here’s the takeaway. Your YouTube audio will never play back as loud as Soundcloud because YouTube’s LUFS normalization system won’t let you. On Soundcloud, your audio is smashing 0 dB, which tells me it was peak normalized. YouTube won’t let you stay at peak, so YouTube will never be as loud. YouTube will turn down your audio however much is required to make it average -14 LUFS.

Takeaway #2 is that over-compressing (meaning dynamics compression, not data compression) the audio to “sound louder” will actually work against you on YouTube. I don’t know if heavy dynamics compression is part of your mixdown process or not. But if it is, you can actually recover loudness on YouTube by reducing the amount of dynamics compression applied. The problem is that heavy compression makes the audio “seem” louder on the LUFS scale, which means it gets turned down more. And when compressed audio gets turned down, everything seems quieter because everything is in the same loudness range. With more headroom (less dynamics compression), your audio won’t get turned down as much, and as a result, actually seem louder when played back on YouTube.

However, if you need heavy compression and 0 dB peak normalization for distribution to other places, and if you like the sound the way it is on Soundcloud, then keep doing what you’re doing and simply remember not to compare YouTube to Soundcloud. Soundcloud lets you post whatever you want without alteration. YouTube will clamp anything that’s too loud. It is what it is.

[Entheo]

wow, that’s fantastic - thank you so much Austin! you’ve just won a free subscription to my YouTube drumming channel (haha!)

i do get very greedy when i mix in Audacity - i generally push to 0 db. on top of that, previously when i did my final A/V mix in Movie Maker, I would slam the volume there too. i guess i should back off the gas!

in your esteemed opinion, does it still make sense to final mix in Shotcut (with your recommended settings) vs. just continuing with WMM at lower volumes? do you think i’ll get better audio?

again, thanks so much for your help! i know my musical craft, but not sound engineering, that’s for sure!

[Helion]

I havn’t paid much attention to youtube audio conversion and compression, but here’s some info
https://support.google.com/youtube/answer/4603579?hl=en

[Entheo]

Thanks so much again, Austin! Not sure if you saw my last reply, but I’d like to bug you with one last question. My current videos produced in WMM have the following codecs, according to YT stats for nerds: avc1.640028 (137) / opus (251). Do you think that with your Shotcut recommended settings that I’ll get appreciably better audio?

Thanks again for all your help!

[Austin]

Wow, this is a tough one.

The main thing is to get YouTube to stream your videos with the Opus codec instead of AAC. AAC will wreck your cymbals. Since you’re already getting Opus when using WMM, it’s probably fair to say that the audio quality difference between Shotcut and WMM would only be apparent to people working in critical listening environments. That’s merely a guess since I haven’t done an A/B test with your WMM videos to prove it.

The only thing we know for sure is that uploading WAV to YouTube will get you the best quality you can get. In that sense, Shotcut empirically wins. Whether there is an audible difference between Shotcut and WMM is yet to be determined, but is likely to be small.

With that in mind… since you’re already familiar with WMM and have an established workflow with it, I could see a case for sticking with WMM. If it ain’t broke, don’t fix it.

On the flip side, Shotcut offers certain advantages that could be useful in the future:

• View > Scopes > Audio Loudness from the Shotcut menu will provide you with audio meters that natively work in LUFS. This will let you know in real-time how much the music is above or below the -14 LUFS threshold while doing your mixdown. For the work I do, this is a super-big deal to me.

• Shotcut offers multi-track video and audio support that is much easier and more powerful than WMM. Graphics and titles and transitions can be fancier, for instance.

• If something happens that it becomes necessary to do the 1440p trick again to get Opus on YouTube, it will be easier to export 1440p with Shotcut than WMM. Shotcut in general offers tons of control for your exported output that WMM does not. If you do your final audio mix in Shotcut, it’s even possible to export just the audio (no video) to a WAV file as your mixdown master. This means you could export audio+video for YouTube and then export WAV audio alone for Soundcloud, from the same Shotcut project, which might save you some workflow steps. It may improve file organization on your hard drive and enforce mix consistency, too.

• If you ever want to listen to your videos for the fun of it on high-quality headphones or speakers, you might appreciate having masters stored on your hard drive with WAV or FLAC or ALAC or even high-bitrate Opus audio, instead of compressed WMM audio. When you have access to the masters and a good listening environment, you will definitely notice a difference between Shotcut and WMM output, which might be valuable to you personally. So, you might want to consider yourself or your children’s future listening preferences in addition to the YouTube audience.

One other thing to consider is that more audio platforms than not will enforce some kind of LUFS normalization. Spotify and iTunes are at -16 LUFS, last I knew. The industry players have really moved away from smashing 0 dB. Their goal is to make all their audio assets the same perceived volume so that customers don’t get annoyed with reaching for the volume control at every new song. This is good not only for customers, but also for mix engineers because it means they are free to allow 16 dB of headroom in their mix without fear of sounding quieter (and therefore less “full”) than everybody else. It removes the competition to be the loudest of the pack. If you eventually expand to more platforms that enforce normalization, then it will benefit you to mix with their LUFS targets in mind. Shotcut can provide you with that information infinitely better than WMM.

That’s my two cents. WMM is fine for today, if you don’t intend to do more than what you’re doing. But if you have any thoughts toward growth in the future, Shotcut can take you much further.

[Entheo]

Again, thanks so much for the detailed reply, Austin. I do have my workflows down when it comes to recording my drum (and piano - a separate workflow) videos. The main thing I want to avoid is going down a technical rabbit hole which distracts from music making (I’ve done that before with a digital studio, which I ended up selling and getting an acoustic piano).

My workflow will probably not change much. If I could easily transition from WMM to Shotcut I would do it. Here’s a question for you - would you be willing to do a lesson or two with me on Google Meet? I would be happy to pay you for your time. I could provide you with my current step-by-step workflow so we could (hopefully) cut to the chase in the lessons. I suspect you could also give me some excellent pointers on mixing, levels, etc.

Please let me know if you’re interested, and thanks again for all your help!

[Austin]

This sounds possible! I’d love to see you get a fast start and not go down a rabbit hole. I know first-hand that musicians need to keep their focus on the music, not the gear.

I’m getting off work very late tonight. I will send a personal message to you tomorrow with some details and see what we can work out.