• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

YouTube now uses VP9 as standard (for popular videos)

What happened to YouTube before?

You may have noticed that since around a year ago, youtube videos look like crap (you may remember this neogaf thread)
The reason is the introduction of the new dynamic adaptive streaming (DASH) formats.
The new formats have drastically lowered the video bitrate (720p: form 2-3 to 1-1.5 1080p: from 3-5.9 to 2-3 MBit/s)

What is happening to Youtube now?

Google has quietly begun to use VP9 as default encode for new popular videos
VP9 is a more efficient codec. The bandwidth used is similar for 720p, and 50% higher for 1080p+.
Here is a 720p F-Zero GX comparison done by Doom9/overclocker forum member Nintendo Maniac 64:

iCxgC72dVYE7R.png
ibjt5PUMqX3n5U.png

iIBej6Qrkmhkb.png
ibqxRK86b2dCPI.png
F-Zero GX video: https://www.youtube.com/watch?v=BLxHjixnf3c
Yellowstone 4k video: https://www.youtube.com/watch?v=Y-Pa8yYXews
In Chrome/Opera, you can right click the video to find out if it uses vp9:
Source: [1] [2]

downsize if old
 

AlucardGV

Banned
i don't really know those things but is vp9 the same thing as h265, or similar? i know there was h264 for lower bitrate but same quality
 

Oppo

Member
If they switched to HEVC (H265), they could have even smaller filesizes and maintain the quality.

Yeah but very few devices support this.

I haven't investigated VP9 properly but H.264 was still superior to VP8 last I checked. H.265 is like future tech for anything but desktops.

It's been a little weird seeing people freak out over webm files over GIFs. H.264 still is better. Of course.

Remember that embedded devices still need specific codec support.
 
i don't really know those things but is vp9 the same thing as h265, or similar? i know there was h264 for lower bitrate but same quality
AVC/h.264 widely popular compression standard (for HDTV/BluRay)
HEVC/h.265 next-gen compression standard in development, 30-50% more efficient (for UHDTV)
VP8: streaming codec, originally developed by On2. Made open source after purchased by Google.
VP9: Googles next-gen open source streaming codec, efficiency somewhat lower than HEVC (some patented techniques had to be avoided)

Is this some kind of tech they got out of the Twitch acquisition?
No, see above
 

mothball

Member
What does one use to encode VP9, anyway? I tried making a VP9 webm and promptly quit because it was encoding at 0.5 fps.
 

TheD

The Detective
Know what would be nice? Them fixing the fucking video player so that it starts playing higher quality video from when you select it and not sometime later in the video/never (and the only to fix that is reload the video, select the right quailty then seek to a place in the video it has not loaded yet).
 
Sorry for being dense, but in the OP, which of the comparison screenshots is VP9 and which is... whatever else is currently in use?
 

Schrade

Member
VP9 is a more efficient codec. The bandwidth used is similar for 720p, and 50% higher for 1080p+.

[Citation Needed]

Please tell me how VP9 is more efficient (i.e. better) than h.264? No talking about patents. I'm only referring to technology. Trying to go by Google's shitty compression settings for h.264 vs. their better settings for VP9 is not admissable.
 
[Citation Needed]

Please tell me how VP9 is more efficient (i.e. better) than h.264? No talking about patents. I'm only referring to technology. Trying to go by Google's shitty compression settings for h.264 vs. their better settings for VP9 is not admissable.

I don't know anything about compression algorithms but here is a comparison.

http://digitalstudio7.blogspot.com/2014/01/x.html

Also heres a 4k VP9 video. Looks real nice.

http://www.reduser.net/forum/showthread.php?111230-Google-VP9-4K-HD-Sample
 

alr1ght

bish gets all the credit :)
Still looks like shit. Can't blame them though, they must go through an insane amount of bandwidth. I'd like to see them offer a premium service for uploaders to show their videos in high quality.
 
[Citation Needed]

Please tell me how VP9 is more efficient (i.e. better) than h.264?.
sure, here is a quick one: http://www.m-hikari.com/ams/ams-2013/ams-137-140-2013/sharabaykoAMS137-140-2013.pdf
While [HEVC] provides 31 percent better compression rates in keyframe-only mode and about 40 percent improvement in intercoding mode compared to [x264], VP9 is only 18 percent better than [x264] in both modes.
Conclusion said:
Both VP9 and HEVC compression standards provide higher compressoin effiency compared to the current industrial video compression standard AVC.
HEVC provides better compression rates than VP9, but VP9 is patent-free and can be used without licensing expenses.

In our experiments we tried to figure out the main reasons for HEVC to be more effcient in intra coding compared to VP9.
We showed that more angular intra prediction modes provide the most significant infuence on intra compression effciency (about 7.7% bitrate savings).
The HEVC SAO postprocessing stage has less impact (about 1.3% bitrate savings).

Finally the modifed HM encoder with 10 intra prediction modes and without SAO post-processing is still about 7% more effcient compared to the VP9 encoder. We assume this is due to the adaptive HEVC entropy coding.
If you referring to this paper: http://iphome.hhi.de/marpe/download/Performance_HEVC_VP9_X264_PCS_2013_preprint.pdf
According to Doom9 they used unfair quantisation settings, making the test rather invalid.
 

Schrade

Member
So from what it looks like.. Google is just allowing for more bitrate and using less extreme compression settings for VP9 videos while they are using extreme compression settings and less bitrate for h.264.
 

Furyous

Member
What's the download size on VP9 look like? I'm asking because I want to see how much space my ripped personal youtube videos take up on my hard drive.
 

Schrade

Member
sure, here is a quick one: http://www.m-hikari.com/ams/ams-2013/ams-137-140-2013/sharabaykoAMS137-140-2013.pdf


If you referring to this paper: http://iphome.hhi.de/marpe/download/Performance_HEVC_VP9_X264_PCS_2013_preprint.pdf
According to Doom9 they used unfair quantisation settings, making the test rather invalid.

From that doc:

For AVC evaluation the JM reference encoder and Elecard Stream Analyzer
were used. It was also configured to work in the “constant quantizer” mode.

Using the absolute worst encoder (reference) in CQ mode is not how you do comparisons such as this.

x264 using CRF will almost always yield better results.

They need to be consistent throughout using x264 in every step of the way with the optimal settings.

I think foxyshadis (from the Doom9.org thread) says it best:

foxyshadis said:
It's amazing that they apparently run x264 on ultrafast/realtime for everything and give it wretched quality, then likely spend hundreds of times as much cpu time on their VP9 encodes that aren't universally compatible. You'd almost think it was intentional, maximizing the differences to sell their format. (The way On2 used to.)

VP9 is a good format, there's no denying that, but if they'd done the same for their h.264 encodes they'd look quite a bit better.
 
Top Bottom