jcjr--Youtube uploads

Share your music, collaborate, and partake in monthly music contests.
RELATED
PRODUCTS

Post

So thanks to advice from an earlier thread, got my feet wet uploading a couple of youtube vids. Premiere Elements 15 if I feed it a tacky 1080p bmp file and an uncompressed 44.1k audio file, defaults to uploading youtube at 1080p audio encoded at AAC 128 kbps, which in the earlier thread was advised is about as good as youtube will do nowadays. It doesn't sound so bad that my old ears can tell considering the bitrate.

Well, on 128 kbps mp3 with older generation encoders (tis been years since I encoded 128 kbps mp3)-- The most obvious problems were watery or chirpy cymbals. Cymbals seem to survive the AAC purt well, but unless my ears are just messing with me right now, the AAC might make the ac guitar tones "watery" sounding compared to lossless flac or 320 kbps mp3.

Or maybe the 128 kbps AAC tends to sound more solid on louder parts of the music and less-solid on less-loud parts of the music? Dunno. Or maybe something else entirely.

https://www.youtube.com/watch?v=RiG4mEV9w5E

https://www.youtube.com/watch?v=VU9UQvA44nM

Post

Did some experimentation, will describe details later. The free program VirtualDub (assuming PC platform) seems possibly better-suited to youtube uploads than Premiere, for the static picture + audio file task.

https://www.youtube.com/watch?v=Dm30k70R08Y

Post

So what I've learned so far, not much. Ya'll probably already know.

The Premiere preset for uploading to youtube seems in accordance with what youtube wants. Premiere claimed to upload 128 kbps 44.1K AAC audio, and immediately after uploading and also some days afterwards the right-click "stats for nerds" on the youtube vid shows that same info. The string is "mp4a.40.2(140)" which when I looked it up apparently means 44.1k 128kbps AAC.

But there was something minor about some features that bugged me about the sound. My ears are old, subject to set, setting and placebo effect, generally not reliable but still, something there...

So I tried VirtualDub which on PC will upload an AVI container with the single uncompressed bmp image file and the 44.1k 16bit uncompressed wav file. This sounded "better" somehow but "stats for nerds" reported "mp4a 40.2(18)" which supposedly means 48k 96kbps AAC. Silly theories why 48k samplerate 96kbps POSSIBLY psychoacoustically sounded better later on... YT also transcoded the 1080p BMP file down to a silly low-res 640 X 360 but I don't really care so much about the image quality.

Thinking, "Aha! If left to its own devices youtube would rather have 48k audio" I did another VirtualDub test, same 1080p image file but a 48k 16 bit uncompressed WAV I samplerate converted locally using "the highest quality available" wondering if maybe youtube doesn't always use the highest quality samplerate conversion. That 48k AVI video also got encoded by youtube exactly the same, audio at 48k samplerate 96kbps AAC.

So yesterday I was about to report the findings on a kvr message, but then noticed that after about a day, youtube got around to re-coding my two virtual dub test uploads. Both the 44.1k and 48k uncompressed audio files, after about a day, got promoted to 44.1k 128kbps AAC, the EXACT SAME encoder settings for both video and audio, as was originally supplied by the Premiere youtube upload preset.

However, so far as my fallible ears can be trusted, I think maybe the youtube encoding quality is somehow a little better than the Premiere AAC encoding at the same bitrate.

So the provisional opinion as of now-- For this kind of static picture music video, youtube really does seem to like 44.1k samplerate, 128kbps AAC and it doesn't sound so bad. Because it eventually transcoded my 48k samplerate uncompressed audio to that 44.1k samplerate (after about a day at 48k 96kbps AAC), seems maybe the best one can do is send it 44.1k 16 or 24 bit uncompressed audio and let YT do whatever the hell it wants with the data! :)

https://youtu.be/qSfo9tJtV-c

Post

YouTube Audio Sample Rate

AAC : 44100 khz
OPUS : 48000 khz

Try to upload your video with 1440p (2560x1440) to get the opus codec
Last edited by clothem on Mon Feb 11, 2019 12:52 pm, edited 1 time in total.

Post

Opus vs AAC | sample rate

Image
Image

Post

JCJR wrote: Mon Feb 11, 2019 5:38 am but "stats for nerds" reported "mp4a 40.2(18)" which supposedly means 48k 96kbps AAC.
I have no idea how those numbers turn into that resolution.
But I get <mp4a.40.2 (140)>, consistently. These were 48khz 24-bit audio rendered with the 1920x1080p video.


To me, for instance, the Final Cut Pro X AAC encode is garbage, so I would prefer Youtube just by ear.

Post

clothem wrote: Mon Feb 11, 2019 12:48 pm YouTube Audio Sample Rate

AAC : 44100 khz
OPUS : 48000 khz

Try to upload your video with 1440p (2560x1440) to get the opus codec
Thanks for the good info, clothem! Might try that sometime.

The 48k bandwidth in your spectrographs certainly looks better. Maybe am remembering wrong but last time I looked even 320kbps mp3 (44.1k samplerate) seemed to brickwall well below 20 kHz so maybe it has as much to do with guardbands for 44.1k samplerate compressed audio rather than bitrate? Dumb ignorant speculation.

Post

jancivil wrote: Mon Feb 11, 2019 6:06 pm
JCJR wrote: Mon Feb 11, 2019 5:38 am but "stats for nerds" reported "mp4a 40.2(18)" which supposedly means 48k 96kbps AAC.
I have no idea how those numbers turn into that resolution.
But I get <mp4a.40.2 (140)>, consistently. These were 48khz 24-bit audio rendered with the 1920x1080p video.

To me, for instance, the Final Cut Pro X AAC encode is garbage, so I would prefer Youtube just by ear.
Thanks jancivil. Over the years occasionally I had to program with audio codecs so can remember a little but not much and of course things keep changing.

I googled for the <mp4a.40.2 (140)> and <mp4a.40.2 (18)> strings and found the same text list several places purporting to explain the formats. Most of the formats in that list are 48k samplerate or related multiples. Only a couple, including that <mp4a.40.2 (140)> are listed as 44.1k samplerate or related multiples.

Most audio for video defaults to 48k or its multiples. I thought it somewhat odd that youtube would seem to like video with 44.1k audio so much. Maybe they like it because they wooed so many big labels into posting entire catalogs, which would typically be 44.1k CD samplerate. That is the main reason I mostly stayed with 44.1k samplerate over the years, as it is the "base samplerate" for high-quality mp3 as well as "CD audio". I didn't want to run in 48k, 96k or 192k and then have to always downsample at the end to make an mp3. Maybe putting the cart before the horse but hey...

The three major samplerate families are multiples based on 32k, 44.1k and 48k. When I would program dealing low-level with "older codecs" each available bitrate in the long list was only encodable if you feed the encoder a specific samplerate. For instance if you ask it to do a bitrate associated with 48k but give it a 44.1k audio stream, it would either say "no can do" or go ahead and encode a file that subsequently will decode and playback at the wrong pitch. So when dealing low-level with the codecs of yesterday, after the user supplies his bitrate pref, then if his audio file is the wrong samplerate, I would automatically samplerate convert to the rate that the codec wanted to see in order to crank out that requested bitrate.

It was the same thang decoding and playing compressed files. If the user's sound hardware or daw project is set to 44.1k but a decode delivers a 48k, 32k, or even 22k or 8k samplerate stream, I had to auto-samplerate convert to make it work as expected.

It is possible to do low- or high-quality samplerate conversions. Even high-quality SRC isn't expensive on modern fast computers, but there were times I seemed to catch an OS or tool picking a low-quality SRC for a task even though there is also high-quality SRC buried down deep in the guts of the code. It just makes me a little paranoid about unnecessary SRC done "out of sight, out of mind, beyond my control." So if youtube would rather be streaming out 44.1k audio, I wouldn't want to give it 48k audio and force youtube to do an "unnecessary" SRC. Or vice-versa. If it had turned out that YT would rather be using 48k audio (as in clothem's OPUS example) then I would rather deliver 48k to YT. Or whatever. Just senile blather anyway. :)

https://youtu.be/fff_6NCb88k

Post

JCJR wrote: Mon Feb 11, 2019 5:38 am
If it had turned out that YT would rather be using 48k audio (as in clothem's OPUS example)
I uploaded my videos with 44.1 not 48 :wink:
If your original audio is 44.1, stay at 44.1 :)

https://support.google.com/youtube/answer/6039860

Post

Virtualdub is pretty good for YouTube audio
because the audio get not re-encoded by virtualdub and 0 normalization on YouTube

But the color is not the same as the original
To keep the same color go to Video > Compression > select UTvideo 420 bt709

The audio quality will be probably compromised, but you can keep the same color on YouTube

Post

Comparison Color

Uncompressed : youtube.com/watch?v=RpWGwzveKa0 (WRONG COLOR)
UTvideo 420 : youtube.com/watch?v=R1geWo-sHRo (ORIGINAL COLOR)

Post

JCJR wrote: Tue Feb 12, 2019 5:53 am
jancivil wrote: Mon Feb 11, 2019 6:06 pm
JCJR wrote: Mon Feb 11, 2019 5:38 am but "stats for nerds" reported "mp4a 40.2(18)" which supposedly means 48k 96kbps AAC.
I have no idea how those numbers turn into that resolution.
But I get <mp4a.40.2 (140)>, consistently. These were 48khz 24-bit audio rendered with the 1920x1080p video.

To me, for instance, the Final Cut Pro X AAC encode is garbage, so I would prefer Youtube just by ear.
Thanks jancivil. Over the years occasionally I had to program with audio codecs so can remember a little but not much and of course things keep changing.

I googled for the <mp4a.40.2 (140)> and <mp4a.40.2 (18)> strings and found the same text list several places purporting to explain the formats. Most of the formats in that list are 48k samplerate or related multiples. Only a couple, including that <mp4a.40.2 (140)> are listed as 44.1k samplerate or related multiples.
I googled for it and only got more info I don't understand. That's what it looks like whether or not it was a 44.1 or 48k audio put in the video. Only more recently have I done 48k, which I'm def not married to.

I don't know what 140 or 18 means there.

I'm finding out basically with audio with not very much headroom below 0dB YT is penalizing me, one peak at -1dB is enough.

Post

Thanks clothem. That YT format link you posted says 24 bit FLAC accepted. Will have to test and find out if VirtualDub can embed a FLAC chunk in an AVI. So far I was loading my lossless FLAC versions then saving temp 16 bit PCM 44.1k uncompressed WAV for inclusion with VirtualDub, superstitious maybe YT would barf on higher res.

I can see the color difference in your examples though it isn't drastically bad to my eye but in my case I'm just hoping my eye will last as long as I do. :) Will test and see what happens.

Am not real concerned with the graphics quality. Maybe one day will get raided by the Graphics Good Taste SWAT team, "Mr Chandler we have you surrounded. Keep your hands where we can see them, put down the mouse and step away from Photoshop!"

YT must have busy and idle times. A couple files I uploaded this weekend came back almost instantly encoded for 44.1k samplerate, 128 kbps. But that "Skunks On Fire" video uploaded Monday, at this moment, is still at that "transitional" state of 48k samplerate, 96 kbps. Maybe the transcode servers are just busier during the weekdays.

Noticed that the Amazon-written default web browser for my old kindle fire (overall it works about as good as any) can't play those 640 X 380, 48k 96kbps videos. It works good on the files that have already got promoted to 44.1k 128kbps. The old fire has fairly good-sounding built-in audio, was one of the first gens that were sorta expensive. Dunno how the modern cheap fire's sound.

OK I uploaded this video only an hour ago. Was originally at 96kbps right after upload but they have already promoted it to 128kbps. Maybe Skunks just got promoted as well.

https://www.youtube.com/watch?v=eA3ScQ2NO4s

Post

jancivil wrote: Tue Feb 12, 2019 7:16 pm I googled for it and only got more info I don't understand. That's what it looks like whether or not it was a 44.1 or 48k audio put in the video. Only more recently have I done 48k, which I'm def not married to.

I don't know what 140 or 18 means there.

I'm finding out basically with audio with not very much headroom below 0dB YT is penalizing me, one peak at -1dB is enough.
Hi jancivil. I don't understand a lot. This is the kind of list I could find:

Code: Select all

139          m4a        audio only DASH audio   49k , m4a_dash container, mp4a.40.5@ 48k (22050Hz)
140          m4a        audio only DASH audio  128k , m4a_dash container, mp4a.40.2@128k (44100Hz)
160          mp4        256x144    DASH video  108k , avc1.4d400b, 30fps, video only
133          mp4        426x240    DASH video  242k , avc1.4d400c, 30fps, video only
134          mp4        640x360    DASH video  634k , avc1.4d401e, 30fps, video only
135          mp4        854x480    DASH video 1155k , avc1.4d4014, 30fps, video only
136          mp4        1280x720   DASH video 2310k , avc1.4d4016, 30fps, video only
137          mp4        1920x1080  DASH video 4748k , avc1.640028, 30fps, video only
17           3gp        176x144    small , mp4v.20.3, mp4a.40.2@ 24k
36           3gp        320x180    small , mp4v.20.3, mp4a.40.2
43           webm       640x360    medium , vp8.0, vorbis@128k
18           mp4        640x360    medium , avc1.42001E, mp4a.40.2@ 96k
22           mp4        1280x720   hd720 , avc1.64001F, mp4a.40.2@192k (best)
Because only the first two specify samplerates, I make risky assumption that the others are based on the 48k samplerate family.

Do you use a true peak meter? I've been using a meter seems quite good, TBProAudio dpMeter, which I see he now shows "discontinued" on his web page though it can still be downloaded. He has a couple of fancier current meters, but the original seems good enough for my uses.

I wrote a little lookahead true peak limiter and adjust it so that dpMeter never indicates higher than -1 dB true peak. From asking around awhile ago, supposedly metering isn't an "exact science" inasmuch as different well-written meters might disagree some small amount, a tenth of a dB or whatever. My homebrew limiter doesn't measure true peaks exactly the same as dpMeter and I usually have to set its ceiling anywhere between -1.1 to -1.3 dB to keep from going over -1 dB true peak in dpMeter. But that's alright, close enough for rock'n'roll for a retired geezer. :)

I found that VirtualDub can't recognize FLAC files. Also, maybe it would build an AVI from a 32bit float WAVE file, but it doesn't know how to audition-play them. So this last song I tried embedding a 44.1k samplerate, 24 bit WAV rather than 16 bit, not that ears could probably hear a difference, and it seemed to upload OK.

https://www.youtube.com/watch?v=Ixe1VLngeJg

Post

Comparison AAC vs OPUS

AAC : youtube.com/watch?v=9iQKHI8W2yg
OPUS : youtube.com/watch?v=ByBFkRL07FE

I can see the difference without problem

Opus win ! ;)

Post Reply

Return to “Music Cafe”