H264 vs H265 for YouTube: Which Codec to use at 1080p, 1440p & 4K

[00:00] h.264 or h.265 which codec should you

[00:04] use for rendering your videos for

[00:06] YouTube to answer this question in

[00:08] today's video we are going to look at

[00:10] the difference in render times and we're

[00:12] also going to look at the actual visual

[00:14] quality of the final product on YouTube

[00:17] when rendered either in h.264 or h.265

[00:21] now the outcome of this analysis might

[00:24] actually surprise you quite a bit

[00:25] because even though h.265 or hevc is

[00:30] definitely the superior codec that

[00:32] produces the same visual quality with a

[00:34] smaller file size this doesn't always

[00:37] translate into the best possible quality

[00:39] on YouTube because this highly depends

[00:42] on the resolution that you're rendering

[00:43] in now before diving right into results

[00:45] for today's video I want to mention that

[00:48] all of the tests I have performed using

[00:49] DaVinci Resolve the free version as

[00:52] since version 17 something something

[00:55] DaVinci Resolve supports rendering using

[00:57] h.265 moreover the I'm using for these

[01:01] tests comes from Modern Warfare version

[01:03] 2 so I've used gaming footage at 60fps

[01:05] however the results that I'm showing in

[01:07] this video should hopefully also be

[01:09] applicable for other type of footage

[01:10] such as Talking Heads GoPro or Drone

[01:13] footage and with having said that let's

[01:16] start with our render times now

[01:18] generally speaking you would expect the

[01:20] render times to actually decrease when

[01:22] going from h.264 to h.265 now while this

[01:26] is generally true for most people it

[01:28] also sort of depends on the type of

[01:30] Hardware that you're using to render

[01:32] your videos you see when you render

[01:34] using h.264 you will use 100 of your CPU

[01:39] or rather all of your threads that you

[01:41] have available to render that video that

[01:43] also translates into extremely high

[01:45] temperatures when for instance you have

[01:47] a 3900k on the other hand if you render

[01:50] using h.265 then most of this work is

[01:54] actually done on the GPU and therefore

[01:56] your CPU isn't really doing too much but

[01:59] then again if you have a beastly CPU

[02:01] like the 3900k with 32 threads then

[02:04] going from that to rendering on your GPU

[02:07] isn't actually going to provide you with

[02:09] much of an improvement in render speeds

[02:12] so for my system you can see the render

[02:14] times that I'm getting for a sample one

[02:16] minute clip without any edits or any

[02:18] overlays or any effects on this graph

[02:21] right now now for this video I'm going

[02:23] to be using these three different

[02:24] resolutions that you see here 1080P

[02:27] 1440p and 2160p or 4K the orange bars on

[02:32] this Plot show the render times for a

[02:35] one minute sample clip of each of these

[02:37] resolutions in h.264 whereas the blue

[02:39] bars represent the same clip but

[02:42] rendered at age 265 as you can see in

[02:45] this very idealized case I actually do

[02:47] get Improvement in render speeds when

[02:49] moving from h264 to h.265 however as I

[02:53] mentioned before this is a very

[02:54] idealized case as there are no cuts no

[02:57] overlays no text no effect Supply to the

[03:00] clips whatsoever and therefore I also

[03:03] like to talk about what I'm seeing when

[03:05] actually rendering a proper clip out and

[03:08] here results are a little bit varied as

[03:10] you can see rendering this project here

[03:12] took roughly 30 seconds longer using

[03:15] h.265 compared to h.264 then again I

[03:18] should definitely mention at this point

[03:20] that these results are highly sensitive

[03:22] to the type of harder that they're using

[03:24] and more specifically they're going to

[03:27] be hugely different if you have a CPU

[03:29] that has much less threads compared to

[03:31] the 3900k for example when I repeat the

[03:35] exact same test on my 7700k you can see

[03:38] that I'm gaining roughly 60 to 70 speed

[03:41] improvements by going from h.264 to

[03:44] h.265

[03:45] by the way natively it's not possible to

[03:48] play back hevc videos in Windows a

[03:51] workaround is to Simply use the VLC

[03:53] Media planner which then allows you to

[03:55] actually play back your hevc video clips

[03:58] however if you're using DaVinci Resolve

[04:00] to edit your videos you are likely still

[04:02] not going to be able to play back your

[04:04] videos as you will get this media

[04:05] offline warning so to fix this go to the

[04:08] Microsoft store search for the hevc

[04:11] video extension pay the one dollar that

[04:14] this costs and install the extension

[04:16] after which you'll see that the video

[04:18] will play back just fine in DaVinci

[04:19] Resolve now if so far you've found any

[04:22] value whatsoever from this video then I

[04:25] would highly appreciate it if you guys

[04:26] could smash that like And subscribe to

[04:28] the channel currently only five and a

[04:30] half percent of people that are watching

[04:32] my content are also subscribed so if

[04:35] you're enjoying these super deep dive

[04:36] videos where I'm showing actual data to

[04:38] prove a point then please consider

[04:40] subscribing it would mean the absolute

[04:42] world to me if you could grow this

[04:44] community even further

[04:45] but let's shift gears now and talk about

[04:48] the actual visual quality of the video

[04:50] after you've uploaded it to YouTube so

[04:53] for this I have performed a bunch of

[04:55] different render tests using h.264 and

[04:58] h.265 at various different resolutions

[05:00] and various different bit rates I

[05:03] uploaded all of these videos onto

[05:05] YouTube then re-downloaded them and

[05:07] analyze them using vmaf or video

[05:10] multi-method assessment Fusion which

[05:12] kind of combines different quality

[05:15] metrics into just one number and it's a

[05:18] very nice and convenient way to compare

[05:20] the quality of two different video

[05:22] sources

[05:23] so on this graph once again the Orange

[05:25] Line represents rendering using h.264

[05:27] and the blue line rendering using h.265

[05:30] you can see the different rendering

[05:32] bitrates on the horizontal axis and the

[05:34] corresponding visual quality on the

[05:36] vertical axis with a higher value

[05:38] representing a better visual quality

[05:41] as expected at 4K h.265 always produces

[05:44] the higher visual quality when compared

[05:46] to h.264 rendered at the same bit rate

[05:49] now quite interestingly you can achieve

[05:52] the exact same visual quality on YouTube

[05:54] with h.265 when decreasing your bit rate

[05:58] by 40 megabits per second of course this

[06:01] relationship isn't completely linear and

[06:03] I'd have to kind of extend the graph

[06:04] both to the left and the right to give

[06:07] you this information for all different

[06:08] kinds of bit rates but generally

[06:10] speaking if you're rendering 4K videos

[06:12] for YouTube you should definitely always

[06:15] stick to h.265 both for the better

[06:17] visual quality on YouTube and

[06:19] potentially lower render times depending

[06:22] on your system stepping down the

[06:24] resolution to 1440p we can see pretty

[06:27] much the same behavior by the way if you

[06:28] look at these graphs and think that

[06:30] these could be very handy in order to

[06:32] figure out the best possible bit rate to

[06:34] render your YouTube videos at to get the

[06:37] highest possible visual quality then you

[06:39] would be just right and that's actually

[06:41] something that I've covered in one of my

[06:43] past videos where I talked about the

[06:45] best bit rates depending on your

[06:46] resolution you can check it out Linked

[06:48] In the card right now and I'm actually

[06:50] also working on updating that video with

[06:53] actually better data and that is

[06:55] actually using 4K footage natively in

[06:58] order to do my tests so returning to

[07:00] this plant we can see that once again

[07:02] h.265 is far superior over h.264 at

[07:06] 1440p and will basically always result

[07:09] in higher visual quality on YouTube now

[07:12] the gap between these two curves

[07:13] slightly decreased compared to 4K we can

[07:16] see that the visual quality of a 70

[07:18] megabits render at h.264 is roughly

[07:22] equivalent to a 45 megabit render at

[07:24] h.265

[07:26] so roughly speaking you can decrease

[07:27] your bit rate by 25 megabits and still

[07:30] have the same visual quality as it had

[07:33] with the higher bit rate in h.264

[07:35] however generally speaking I would

[07:37] always recommend to increase your bit

[07:39] rate as far as possible in order to be

[07:41] able to make use of the much better

[07:44] looking video quality on YouTube when

[07:46] rendering at an even higher bit rate

[07:48] finally let's decrease our resolution

[07:50] once again to 1080P and here results

[07:53] look

[07:55] completely different I was actually

[07:57] getting this graph and was very confused

[08:00] till h.265 was far far worse than h.264

[08:06] by a pretty significant margin actually

[08:09] so I was a bit skeptical and therefore

[08:11] decided to re-record a one minute

[08:14] segment in Warzone 2 once again at 1080p

[08:17] rendered it at different bit rates in

[08:19] h264 and h265 uploaded it to YouTube

[08:22] redownloaded everything and analyzed it

[08:24] and that's what happened so once again

[08:27] definitely not what I would have

[08:28] expected so I re-rendered the h.265 once

[08:32] again the exact same clip uploaded again

[08:35] and then it looked like this

[08:37] so essentially what I'm thinking happens

[08:39] is that the much more compressed h.265

[08:41] video Simply doesn't contain enough

[08:44] information to provide a good h.264

[08:47] re-encode on YouTube also this isn't

[08:49] just a random outcome of the vmaf

[08:51] computation as when I visually compare

[08:54] the h.264 render to the h.265 render

[08:57] after downloading the clips from YouTube

[08:59] you can see that the h.264 is actually

[09:02] vastly Superior in terms of visual

[09:04] Clarity compared to the h.265 render now

[09:08] frankly I find it a bit concerning that

[09:10] at 1080p it seems to be a little bit

[09:13] random as to whether you're getting good

[09:15] or bad quality on YouTube when you

[09:17] upload your h.265 videos in 1080P so if

[09:21] having said all of this my conclusion

[09:22] would be that if you have to render your

[09:25] videos in 1080P then I would actually

[09:28] not recommend to use h.265 and use h.264

[09:32] instead however instead of actually

[09:35] uploading 1080 footage you should really

[09:37] consider to upscale your footage to

[09:39] 1440p and render it in that resolution

[09:42] in order to make sure that you can

[09:44] actually make use of the vp9 codec on

[09:47] YouTube now I'll make sure to include

[09:49] the visual quality that you can expect

[09:50] when upscaling a 1080p clip to 1440p or

[09:54] even 4K in this video where I'll talk

[09:57] about the best possible bitrate to use

[09:58] when rendering videos for YouTube in

[10:01] different resolutions so I hope you got

[10:03] value from this video thank you so much

[10:04] for watching have a wonderful day and

[10:06] I'll see you guys in the next video