Hello Sup Forumsents. After some testing using the most recent x264 and x265 encoders I could find, I have concluded that we are now ready to adopt 10-bit HEVC as the standard video codec. It's now only ~3X slower to encode and compression efficiency is ~50% better compared to 10-bit H264.
I transcoded a 720p HQ version of Big Buck Bunny with 10-bit HEVC at a fixed 800 Kbps ABR (1-pass) and with 10-bit H264 at different bitrates (also 1-pass ABR) to compare. Those bitrates include 800, 1144, 1334, and 1600 Kbps.
Anyway I made this thread because I want to know what you all think. Will you now use 10-bit HEVC because of how much it has improved or would you rather stick with old 10-bit H264 for a little longer?
Finally I realize that comparing video quality isn't easy so I made a straw poll for comparing the video files I uploaded: strawpoll.me/10320687
The point of the poll is to confirm the 50% better compression efficiency I estimate 10-bit HEVC has compared to 10-bit H264. If the 1144 Kbps 10-bit H264 encode wins then that means then the compression efficiency drops to 40%. If the 1334 Kbps 10-bit H264 encode wins then that means the compression efficiency drops to 30%.
All the bitrates I mentioned only include video bitrate. All video files have a 96 VBR Opus audi file btw.
Your HEVC encodes look like shit compared to a properly done x264 encode. The truth hurts, son, and x265 just isn't there yet.
If it's good enough for you, whatever, but it's got a long loooooonnnnnnnnnnnggggggggggg way to go before it can match the actual visual quality of x264.
Logan Cox
And you have a shitty laptop too, for the record.
>AMD >never, ever
Austin Clark
>using Fahrenheit for temps >not using Celsius which is what everybody uses >trying to be different >you fucking fail
Hudson Collins
>Your HEVC encodes look like shit compared to a properly done x264 encode. There's only one, the rest are 10-bit H264 encodes. My aim was not to make a super high quality HEVC encode though, I just wanted to compared 10-bit HEVC against 10-bit H264.
>The truth hurts, son, and x265 just isn't there yet. Elaborate. I'm getting ~8FPS encoding 720 on my laptop. Someone with say a desktop i5/FX-83XX would get like 30FPS encoding 720p with the fast preset.
>If it's good enough for you, whatever, but it's got a long loooooonnnnnnnnnnnggggggggggg way to go before it can match the actual visual quality of x264. wuuuuuuuuuut. Have you even seen the encodes I did and compared them? The 800 Kbps 10-bit H264 encode looks like shit compared to the 800 Kbps 10-bit HEVC encode.
Wyatt Rivera
>And you have a shitty laptop too, for the record. Not really, APUs on laptops are okay. They have good iGPUs which help with things like hardware acceleration (ie better video playback).
Gavin Lopez
>looks like shit
I did say "properly done x264 encode" which you're not actually doing.
Besides, you're using lossy source material to begin with, you moron. Get the lossless version of Big Buck Bunny and redo things.
>can't even do proper testing methodology >expects people to consider the results worthy of consideration >"check it out, I can take a low bitrate lossy file and make it even fucking smaller after it's been re-encoded 4x with lossy compression each time..." >I mean really
Gabriel Torres
Playback is never a concern anymore, you're doing video ENCODING not decoding - any modern processor can decode things without issues, even your stupid 10-bit bullshit which is just a gimmick overall and only useful for animated flicks.
>idiots doing 10-bit >never fails
Matthew Carter
Get the raw footage of "Tears Of Steel" and then do an encode and we'll give a shit:
It's only 4TB of raw footage, surely you can crunch that down to a 10-bit file that's what, 125MB ? With great quality, right?
Right?
Henry Campbell
How's vp9 holding up to all this?
Christian Morgan
>I did say "properly done x264 encode" which you're not actually doing. Like I said I just wanted to compare the video codecs senpai. If I was going for quality I would have used CRF but the values on HEVC are nothing like that on H264 so you can't really compare them (ie HEVC 22 CRF does not equal H264 22 CRF).
>"check it out, I can take a low bitrate lossy file and make it even fucking smaller after it's been re-encoded 4x with lossy compression each time..." Sorry best I could do anyway my aim was the above. I could have downloaded the lossless version of big buck bunny but it would have not mattered in the end since the source remained the same throughout the test. All encodes came from the same ~5Mbps 8-bit H264 encode I downloaded. Also 5Mbps seems like a lot for an animation will less motion than a typical action movie desu.
Cameron Hughes
I'm not sure. I kinda gave up when a 720p encoding test on my laptop was at like 2FPS (-speed 1). That was months ago though. Not sure how much VP9 has improved since then.
Josiah King
>Playback is never a concern anymore, you're doing video ENCODING not decoding - any modern processor can decode things without issues, even your stupid 10-bit bullshit which is just a gimmick overall and only useful for animated flicks. x264.nl/x264/10bit_02-ateme-why_does_10bit_save_bandwidth.pdf
10-bit encoders are not just beneficial for chinese cartoons you know.
Juan Cox
Here are some of my results using Handbrake x264 CQ20 vs x265 CQ20. These are all 480p taken from DVD. Not too bad.
King of the Hill Season 1 (13 ~22 min episodes) x264 4.59GB x265 1.8GB MASH Season 1 (24 ~26 min episodes) x264 10.6GB x265 3.98GB X-Files Season 1 (24 ~44 min episodes) x264 12.8GB x265 4.58GB
Kevin Walker
>Handbrake babbys first encoder >8-bit encoders too Also you've made a fatal mistake thinking CQ aka CRF are the same on H264 and HEVC. They're not.
James Campbell
Thank you resident autist. I was pointing out that ~50% sounds entirely reasonable.
Angel Allen
Here, OP. Some reencoded shots using HEVC on ffmpeg. I don't remember the settings, but it did took a while. The filesize reduction however, was pretty good. That being said, it was on a 3570k, so something with more cores could definitely do it faster.
Nothing wrong with HandBrake, it gets the job just fine and can handle all the command line options for x264/x265 without any issues whatsoever.
I know x265 has a purpose, but for my own uses it's just not compatible at this point in time. When it can encode just as fast as x264 can (getting there but not there yet) AND THEN provide the same or better visual quality as well as smaller file sizes (50% would be awesome but I don't really give a shit in the long run) then I'll consider it.
So far in my testing I get better results with visual quality using x264 and that's not going to change anytime soon, and it gets done 5x faster as well.
Adam Kelly
>anime screenshots >like that was a surprise
Ayden Anderson
I guess but what you've done is made HEVC encodes with lower quality than the H264 encodes by using the same CRF. You dun goofed son.
The HEVC video turned out to be almost 70% of the original video's filesize at 240MB compared to 343MB . A 30% decrease in size is no joke, it would have been great to have 7TB of animu instead of 10TB.
Alexander James
Only reason I'm doing x265 encodes is because my stuff including PC will be in temperature controlled storage for the next few months. I'm going from 12TB to 64GB on a RPi3. The more music and tv shows I can shoehorn onto this thing the better. >mfw I will be living in someones closet dining on chicken flavored noodles for 100+ days
Tyler Fisher
>Nothing wrong with HandBrake, it gets the job just fine and can handle all the command line options for x264/x265 without any issues whatsoever. Except the permanently fused x264 and x264 encoders on Handbrake are 8-bit.
>I know x265 has a purpose, but for my own uses it's just not compatible at this point in time. When it can encode just as fast as x264 can (getting there but not there yet) That's never going to happen because of how computationally complex HEVC is. There are some things you just can't magically make less computationally expensive. However despite that you can now achieve about 30FPS encoding 720p video on an i5/FX-83XX like I said. Did you fall for the laptop meme too? Don't be shy I did too.
>AND THEN provide the same or better visual quality as well as smaller file sizes (50% would be awesome but I don't really give a shit in the long run) then I'll consider it. That's partly why I made this thread. I think 10-bit HEVC is now 50% more efficient than 10-bit H264. I want people to look at the test files and tell me whether they think that's true or not.
>So far in my testing I get better results with visual quality using x264 and that's not going to change anytime soon, and it gets done 5x faster as well. Well you should really be using the latest x265 encoder like I am instead of using a stale year old encoder fused into the Handbrake encoder.
Look, download MeGUI and the latest x265 encoder from snowfag (link is at start of thread). You'll change your mind about HEVC, trust me.
Ayden Flores
Check out the latest 10-bit HEVC encoder from snowfag then. I'm pretty sure HEVC just reached 50% better compression efficiency over 10-bit H264.
Download the samples and view them side by side to see for yourself.
Christopher Hughes
>>mfw I will be living in someones closet dining on chicken flavored noodles for 100+ days Sounds pretty bad. Did a "pharmaceutical" deal go wrong?
Carson Turner
>Well you should really be using the latest x265 encoder like I am instead of using a stale year old encoder fused into the Handbrake encoder.
That's your opinion, not a fact, and while you're entitled to have an opinion that doesn't translate into a right to express it, go figure.
>try harder next time, son
Grayson Morris
>we Who is "we"?
Jaxson Scott
Well maybe I should have been specific but anyone with a good desktop processor I guess. I'm waiting for Zen but might build an FM2+ desktop to start encoding in 10-bit HEVC sooner.
x264 was developed by videolan, x265.org is multicoreware, did they just copy the name?
Gabriel Taylor
>Windows
Entire thread disregarded.
Nolan Perez
>x265.org/ This is the x265 encoder developed by MulticoreWare and the one OP is using (snowfag just compiles the windows executable).
>videolan.org/developers/x265.html This is the x265 encoder developed by videolan which doesn't seem to have 10-bit encoding support and is rarely used.
>x264 was developed by videolan, x265.org is multicoreware, did they just copy the name? No x265 is still respected as open source by multicoreware.
Aaron Jackson
At least he's not using Handbrake.
Alexander Brooks
>lincuck Opinion disregarded :^)
Gabriel Peterson
>encoding thread without daiz post the summoning
Ryan Rogers
Fucking daiz, now my phone overheats playing that 10-bit meme shit. It does look good in dark scenes though.
Lincoln Adams
lel you just mad you can't use megui without WINE shitting itself every 5 minutes.
Jayden Nguyen
But x264 and x265 are programs that encode H264/H265, right? So there's now two encoders named x265 by different developers. Seems kind of confusing.
Bentley Lee
I might be wrong but I believe they are the same, only the licensing terms change. x265 was developed by multicoreware and I'm not aware of any other project with the same name.
Sorry mate but it sounds to me like you don't know shit about encode testing methods. When comparing encoding qualities the source being lossy is almost completely irrelevant (what will be supplied to the encoder will be identical) and you want to use a low constant/average bitrate in order to the better quality encoder to be more pronounced.
You're trying to say this should be tested like converting FLAC to 320kbit mp3 vs aac where it will be hard to spot a difference either way when why not rather convert 320kbit mp3 to 64kbit mp3 vs aac. Neither of them will be able to match the source because they are simply both too bitrate starved, but one will still do a better job than the other.
Hunter Foster
Can anyone explain the h264 vs hevc crf problem? Why didn't it stay the same, why are tye crf values on hevc different than the ones on h264?
Sebastian Sanchez
Fine.
Ethan Anderson
You fool, what have you done?! Now he'll force 10-bit hevc encodes and make people's phones burst into flames. The end of days is upon us.
Nathan Lee
default crf is 23 for h264 and 28 for hevc. Given this I think hevc is just limited in how "shitty" in can look.
Jackson Nguyen
HEVC does not respect freedom. A patent mess.
Colton Garcia
H264 doesn't respect your freedom either. Pirates and people still use it for personal use.
Wyatt Gonzalez
Why should video pirates care, again?
Eli Ross
Isn't VP going through the same shit?
Evan Wright
>It's now only ~300% slower to encode and compression efficiency is ~50% better compared to 10-bit H264
No, HEVC is still not ready. Decoding 8-bit HEVC is still demanding for some CPUs, let alone 10-bit. We need to wait 2-3 years until HW decoding becomes widespread.
Nicholas Davis
Yep, you have to pay royalties if you distribute HEVC-encoded content , just like MP3. That killed H.265 for me.
Blake Evans
Because you're supposed to promote free codecs so it they can have widespread software and hardware support.
Nathan Roberts
You need the industry support to get widespead HW/SW support. That needs money
Chase Price
This. There was someone posting a korean 60fps UHD video on google drive, but I can't seem to find the link. If you can play that, then you're good to go. The problem is that 99% of computers can't play it without some sort of hiccup.
Lucas Cruz
I recorded a 20Mbps 4K HEVC demo broadcast straight from the satellite. It looked great but my CPU also couldn't keep decoding it at 60fps. However my GPU could decode an AVC transcode of the clip at rock-solid 60fps(thanks DXVA)
Joshua Robinson
Pirates never pay that royalty though.
Jaxon Edwards
>How's vp9 holding up to all this? Not sure, but Daala has has 10- 12- and 14-bit encoding for a while now....
It works with Skylake iGPUs only if I read the source correct
Logan Scott
No, but the content industry will.
Julian Young
I'm back. Sorry I was gone for so long, something bad happened. It's all good now, thank you for keeping the thread alive everyone, glad to see you're all interested in HEVC.
Landon Cruz
Shit, I miss playing RS.
Camden Powell
>No, HEVC is still not ready. Decoding 8-bit HEVC is still demanding for some CPUs, let alone 10-bit. For phones, sure. Laptop i3s/A8s will decode up to 1080p60FPS just fine. Desktops Can probably handle up to 4k60FPS if they have an i5/FX-83XX.
>We need to wait 2-3 years until HW decoding becomes widespread. This isn't really a problem for for laptops though HW decoding for phones would do wonders for battery life and temps.
Ayden Ortiz
>Desktops Can probably handle up to 4k60FPS if they have an i5/FX-83XX. mpv shits on itself when I tried that Korean rip. My computer isn't that much of a slouch since it has a 3570k at 4.3GHz and a 970.
John Watson
>mpv shits on itself when I tried that Korean rip. My computer isn't that much of a slouch since it has a 3570k at 4.3GHz and a 970. Really? Have you tried VLC or MPC-HC? Is playback still problematic with those?
It's an 8-bit HEVC encode too right?
Jaxson Bailey
Get and external hdd?
Aiden Moore
10 bits? Are you serious. Last I checked, everyone has a 24 bit display. Why you want to crap out the colors that way?
Sebastian Lopez
see
Carson Ross
...
Oliver Ramirez
>replying to trolls
Anyway, x265 is still not ready, unless you encode at stupid bitrates like 1000 Kbps like you did here. It's almost universally agreed that once you turn up the bitrate and try to maintain fine detail x265 at its current stage falls apart.
This graph is meaningless. vomit used the 1.9 x265 encoder ya dingus.
Chase Martinez
>It's almost universally agreed that once you turn up the bitrate and try to maintain fine detail x265 at its current stage falls apart. Which is why the tuning is set to none as default. The grain tune in x264 and x265 are not the same and obviously won't behave the same.
Also who the fuck uses a bitrate to encode stuff? You're only supposed to set a CRF and choose a preset. That's it .
Andrew Foster
You use CRF to get an encode with good transparency, which tends to have similar bitrates for similar types of content (grainy movies, etc). In the end you want a stream that's at least 15-20 Mbps for 1080p, 50-60Mbps for 4K for anywhere close to transparency. When you set the CRFs so that HEVC gets bitrates close to that (so for example, CRF 18 for HEVC and CRF 13 for H.264), it falls apart in detail retention compared to H.264 at the same size.
Of course, if you only consume media at SD or HD bitrates for full HD or 4K content, HEVC wins without a doubt. But that's just polishing a shit turd.
UHD BD's will be 80-100 Mbps HEVC, so that's a guideline as to what truly transparent will mean for HEVC encodes.
