You are not logged in. Please login or register.
SmoothVideo Project → Using SVP → Convert Videos to 60fps in just a few clicks
Yay, DASH Manifest now works. And yes, some videos are now only available in lower quality than they were previously. Video formats change all the time. So there are files I have locally that I don't want to replace.
Chainik, for downloading, since you're already using Python, you might as well use YouTube-dl's code which is in the Public Domain according to their license
For many of the low-quality videos, their best version appear to be the VP8 WebM non-DASH version. Sometimes it's the only one with a decent file size.
At least for some of them. Videos <= 480p are downloaded in all kinds of ways
- WebM VP8 non-DASH
- WebM VP9 + 256kb ACC into MKV
- MP4 file with 256kb ACC
With this one, for example, WebM VP8 is the best available
https://www.youtube.com/watch?v=_4WZeBcEvZI
Even better than non-DASH 360p h.264, aka fmt18?
Also, quick question - your 35% size calculation for determining h.264 vs VP9 - is this only for the DASH video stream alone without any audio?
Speaking of audio, since I've now got that Opus mod I rigged up for CompleteYouTubeSaver, I can finally do the quality comparison between 160kbps Opus, 128kbps Vorbis, and 256kbps AAC. Also, it must be noted that Opus does not support 44.1KHz so anything of that sampling rate will be resampled to 48KHz - this is something I will definitely be investigating quality-wise.
I'd be surprised if 160kb Opus compared to 256kb ACC. The main advantage of Opus is that it has a very low latency for live streaming.
For WebM, I calculated 35% improvement for VP9. I'm currently applying the same ratio for VP8 but don't really know the difference. In any cases, the WebM stream I posted above is considerably larger than the MP4 stream.
After doing all these improvements, you can download the latest version here
http://spiritualselftransformation.com/ … 2-beta.exe
With the mass-downloader, if you scan and update the whole playlist and scan again, it still gives several false positive of "better versions available", for various reasons.
You should probably not use the same ratio for VP8 since it's quality is not even better than h.264 at the same bitrate (maybe against h.264 baseline)
And regarding Opus, it actually is flat-out the best general lossy audio codec currently available. Now since YouTube's AAC encoder isn't all that great, it's still possible that Opus could in fact be better.
OK I'll remove the 35% bonus for non-DASH WebM.
Make some tests with Opus and let me know. Do you have good headphones or speakers to test it?
OK I'll remove the 35% bonus for non-DASH WebM.
Well it's still possible it could be better - I'm just saying that you should not blindly use 35% since VP9 is quite a bit better than VP8 and therefore do some testing of the quality difference.
Make some tests with Opus and let me know. Do you have good headphones or speakers to test it?
Let me put it this way, I could hear a difference during an ABX-ing between 128kbps Vorbis and 192kbps Vorbis from YouTube (when the latter format was available anyway).
I don't think the VP8 difference matters all that much. It's only for a few low-quality videos.
As for Opus, looking at this graph, it appears it has great benefits in low bitrates, and that at 128kb, the various codecs even out.
https://www.opus-codec.org/comparison/
Considering the fact that it offers 50kb and 70kb formats, it most likely was added for Live Streaming.
at 128kb, the various codecs even out.
That's why I was making a point about being able to ABX 192kbps Vorbis from 128kbps Vorbis. 
For reference the source video (which I personally uploaded) that was uploaded to YouTube had lossless audio, and I don't just mean lossless-edited audio - the actual audio track was a song that was sourced in FLAC.
Speaking of which, fun fact: YouTube supports FLAC audio in the MKV container. 
Yeah I bet there's a good difference between 128kbps Vorbis and 192kbps Vorbis. I have Beyer Dynamics DT 770 Pro headphones with a pre-amp so I can easily hear the audio differences.
YouTube can read most formats, that's the easy part. It won't support playing it back in FLAC 
Plain and simple, 160kbps Opus is better than 256kbps AAC - no ifs, ands,or buts.
During my ABX-ing, if I by chance played the AAC version first (note that I always play the first track for 5-10 minutes straight), when I switched to the second audio track I could tell literally within 5-10 seconds that it was higiher quality, and every single time this occurred the second and higher quality-sounding track was Opus. It didn't matter if the source audio that was uploaded to YouTube was 22050Hz, 32KHz, 96KHz (all in lossless), or even a 44.1KHz 160kbps MP3 that was 7 years old - Opus was always better.
Interestingly enough, if I ended up playing the Opus version first and then played the AAC version second, it wasn't obvious at all that it was lower quality. I didn't figure out until late that, whenever I did not notice an almost obvious difference within 5-10 seconds, it meant that the first track I played was Opus.
One distinct thing I noticed repeatedly was that the Opus version was constantly more "musical" and enjoyable while the AAC version was almost more "artificial" and boring, ironically even for synthesized electronic songs!
Fun fact: the quality difference between 256kbps AAC and 128kbps vorbis is similar to the difference between 160kbps Opus and 256kbps AAC.
EDIT: Oh, and regarding my audio setup, it's much more budget-fi yet it still amazes me that it seems to still get the job done with results comparable to more expensive setups. It's the Monoprice 9963 IEMs with LostEarBuds single-flange medium-sized tips plugged directly into a Xonar DS with an LME49990 op amp; audio for the ABX-ing was played back in foobar2000 via ASIO (the only way to avoid resampling on Asus sound cards - yes the stupid drivers even resample WASAPI exclusive!)
EDIT 2: I must admit though, I actually didn't do a lot of testing with 44.1KHz native content since it seemed that most of my YouTube videos with 44.1KHz source content did not have Opus encodings (that above-mentioned 160kbps MP3 being the only major one), so I may just test that a bit more sometime tomorrow...other than that though, Opus is definitely better
I did some tests with a 44100hz audio.
Opus 160kb sounds a little bit more natural than AAC 256kb, but also loses much more details. It's kind of a more sharpened picture where you have the edges clear but the details missing. AAC is much richer in details so I prefer 256kb ACC by far.
The sharpening is because AAC consistently has more treble - I noticed this back before they introduced Opus.
However, regarding actual detail, from the minor testing I did I fould that AAC actually had muddier "background" detail, if that makes any sense in an audio context. This was consistent across all sampling rates, but was least noticeable with 44.1KHz content - either that or it was because, out of everything I tested, I was the least familiar with the 44.1khz content (I'm hoping to change that for tomorrow!)
EDIT: And one thing I'd just like to point out, for 32KHz the difference between Opus and AAC was like night and day. It's possible that the resampling has something to do with it since 48KHz is exactly 1.5x of 32KHz while 44.1KHz is some wonky fraction (UPDATE: just checked, it's 1.378125x of 32KHz).
Can you send me the exact video you're testing with? I'll test with the headphones tomorrow.
Here you go, one 32KHz song coming right up:
https://www.youtube.com/watch?v=mu9qC6pxIk0
Here's everything else I tested if you care (save for the 48KHz and higher stuff); note that everything other than the one MP3-sourced sound was sourced from, edited as, and uploaded with fully lossless audio:
22050Hz:
https://www.youtube.com/watch?v=kEGVzY_dtts
https://www.youtube.com/watch?v=X6pAOGDjJLE
https://www.youtube.com/watch?v=pf43SMzTxbw
44100Hz:
https://www.youtube.com/watch?v=oZhCOn5FcUo
https://www.youtube.com/watch?v=Rso1MZz6UuU
https://www.youtube.com/watch?v=Rso1MZz6UuU
https://www.youtube.com/watch?v=xv1ilocUQs4 (160kbps MP3 source!)
The first 44100hz sample you posted is very good for testing because it has a wide range of frequencies: bass, middle and treble.
Format-wise, Opus maintains more the character of the original audio. AAC's bass sounds a bit more square while the treble is a bit higher, with a bit more noise in the background. Also we have to consider that most videos are uploaded to YouTube in AAC format, so I suppose it's easier to re-convert AAC to AAC than AAC to Opus.
Bitrate-wise, however, the 256kb AAC still has much more details and a richer sound than the Opus. But the details are something that most people probably don't notice. I find VP9 and Opus to keep more of the details that people notice, and to drop the details and noise.
That's testing with professional headphones that have a high accuracy reproduction of the whole audio spectrum, and a pre-amp plugged via USB that reduces the analog background noise.
I uploaded the 256kb AAC and 160kb Opus so that anyone can compare easily (the Opus is hard to grab from YouTube!)
https://mega.co.nz/#!XRA2TIKL!zvJNrFgVM … Pw_qpK0COw
https://mega.co.nz/#!nVIWwKZS!UPJa0MVug … 93rTdOinl0
EDIT: Wait! I was doing the tests with VLC. If I play in Windows Media Player, however, Opus sounds better! VLC's Opus decoder is bad.
In WMP or MPC-HC, I would say this file plays better in 128kb Opus.
Here are more audios to compare
32hz. Opus clearly sounds much better
https://mega.co.nz/#!nI5SVCBK!8U6c0hoK1 … j9X3kLoNJQ
https://mega.co.nz/#!6AhzGLQZ!PKV4JJgPR … s3Ol3yz82w
Real-life scenario. Inna - More Than Friends. It's a close call, but AAC sounds richer.
https://mega.co.nz/#!fJABmRSB!gogKytyR0 … ReVDWvavaE
https://mega.co.nz/#!rQJQXaiT!XoyhOwLUj … 4-uFY-rd50
Real-life scenario. Michel Teló - Ai Se Eu Te Pego
https://mega.co.nz/#!2IBWlBYT!wSvNlT3nZ … IODPygjops
https://mega.co.nz/#!KR5mFJ7C!XVyT2llIP … fT5bFwnKQw
The first 2 audios are pure electronic sounds with no background details so perhaps those are better encoded with Opus. For live recordings, however, it's more difficult to calculate the shape (Opus) and storing the pixelated details (AAC) still gives a richer sound.
For Michel Telo, the very beginning with simple audio sounds considerably better in Opus, and the rest of the music with complex sounds sounds better with AAC.
Opus works best for simple sounds or electronic sounds. AAC works best for complex sounds.
So I'll stick with AAC for now.
Mystery
plugged via USB
I hope it's something like this one:
http://www.ebay.com/itm/OYAIDE-Continen … 1652884707
The cheapest decent pre-amp available
http://www.amazon.com/Art-USBDUALPREPS- … uelpre+art
Unlike VP9 that stores more data per byte, I don't think Opus stores any more data than AAC. It simply is more intelligent as to which data it discards.
Again though, YouTube's AAC encoder is quite sub-par compared to non-open source encoders (like Nero AAC).
Also, regarding VLC and WMP, it's likely that WMP is using Wasapi so the 48khz is being played without resampling while in VLC it's being run through DirectSound so it's being resampled by Windows to whatever the "Default format" is set to in Control Panel -> Sound (commonly 44100hz). Now consider that, on Windows 7, the resampling algorithm used is actually really crappy due to a bug and so there's a hotfix for it (alternatively setting 96000Hz seems to work around the issue).
Lastly, you're right about Opus being difficult to retrieve - I can't even get Opus now from that 32khz example I linked to! Update: Oh, it's because my Opus mod for CYS got over-ridden via auto-update by the current version. Guess I should change the version number to 5.6.26.1 or something.
Nintendo Maniac 64
I can't even get Opus now from that 32khz example I linked to.
That one - https://www.youtube.com/watch?v=mu9qC6pxIk0 ?
I can see and play formats 249,250,251 with SVPtube right now.
See my update.
This's because you're not using SVPtube 
Does SVPtube support actual downloading and not just streaming? For proper ABX-ing I need both AAC and Opus downloaded.
The first 2 audios are pure electronic sounds with no background details so perhaps those are better encoded with Opus. For live recordings, however, it's more difficult to calculate the shape (Opus) and storing the pixelated details (AAC) still gives a richer sound.
By electronic, do you mean the equivalent of a direct recording, electronic as in synthesized, or do you mean electronic as in the actual genre? If the latter, well, that's what happens when the majority of my test content (my own youtube "videos") don't have Opus encodings!
Nintendo Maniac 64
Does SVPtube support actual downloading
When you feed ffmpeg with two streams it downloads and merges them...
I won't believe you didn't read SVPtube's "readme"
SmoothVideo Project → Using SVP → Convert Videos to 60fps in just a few clicks
Powered by PunBB, supported by Informer Technologies, Inc.