Need best render settings for 4K video for YouTube?

Hi Guys,

Can anyone tell me the best render settings for 4K video for YouTube?

Thanks in advance,

Zia

This is the YouTube recommendations for uploads which may help.
https://support.google.com/youtube/answer/1722171?hl=en-GB

I export my 1080 material to DNxHD.

I've heard from others that DNxHR also works.

The one caveat is that the DNxHD exports from Resolve do not work, whereas the same encoding settings from Premiere Pro do work. ('Sup with that BMD?) That same may or may not apply to DNxHR.

Thank you Charles and Jim!

Export dnxhr and then use ffmpeg to convert it to h.264 or h.265 for delivery/upload.

Not a great idea, since YT will recompress it.

Not a great idea, since YT will recompress it.

Unless you're on a blisteringly fast connection its the only realistic option. DNxHR SQ (minimum acceptable imo) is 4.3 GB/min at 4k 30p.

If you use a constant quality setting of like 18 (visually lossless)and h.265, the file will be an order of magnitude smaller or more, and you won't be able to tell the difference between it and DNxHR once YouTube gets done destroying it.

Basically it doesn't really matter what codec you upload as long as its high quality footage, as YouTube destroys them all equally.

Not a great idea, since YT will recompress it.

YT ALWAYS re-ecndoes, so you want to feed it with something decent.

YT ALWAYS re-ecndoes, so you want to feed it with something decent.

I agree 100%
Can you give an idea of decent H264 related to frame size and frame rate.
YouTube recommandations seems not that good
Thanks

Yep, youtube recommendations are not good enough. Bitrates are to low.

If you use x264 (in handbrake or ffmpeg etc) use CRF mode with value of e.g. 15 and this will adjust bitrate as needed depending on the source nature. With this setting you can end up with e.g. 15Mbit for high quality very clean HD source or even 70Mbit for crazy noisy source. Idea is that quality remains good regardless of source nature is which key driver for bitrate. Recommending single bitrate value is not good (or I would have to say e.g. 50Mbit which is overkill for many HD sources) as once master looks really good at 15mbit and another needs even mentioned 70mbit. CRF is very good mode of making masters for youtube. Another issue is your upload speed- if it's decent then you an definitely use CRF mode. If you need to squeeze a lot into small file size you may want 2pass VBR with average bitrate chosen based of realistic file size (due to upload speed).

But am I right that you will have compress from Resolve using a very high quality codec to supply a file to handbrake?

Yep, use ProRes or DNxHR or Cineform.

this is mainly ffmpeg related, but in general it should help people understand the various encoding schemes regardless of what application they use.

http://slhck.info/video/2017/03/01/rate-control.html
http://slhck.info/video/2017/02/24/crf-guide.html

https://trac.ffmpeg.org/wiki/Encode/H.264#crf
https://trac.ffmpeg.org/wiki/Encode/H.265

Mentioned values of CRF at e.g. 18 are bit not good enough when you create a "master" which is going to be re-encoded further. I would say you need them bit lower.

Yep, use ProRes or DNxHR or Cineform.

That's just an extra step that I think kind of sucks.

That would be a very good reason for the next version of Resolve to have the option to export directly to ffmpeg/ffmbc.
I would think that a module that just point the output to ffmpeg should not be to difficult to implement. The user could just enter all ffmpeg's parameter for the final render.
Personally I would also like the option to export a fully compliant bluray file from the deliver page.
All this would not only add more flexibility to Resolve but also save any re-encoding.

Resolve has atm. no export plugin API, so this is not going to happen any time soon (if ever).

That's just an extra step that I think kind of sucks.

People who are serious about their work or just want the best results use the best tool for the job rather than complain that some app is not an AIO Swiss army knife. Those tools tend to be substandard consumer versions dumbed down for the masses anyway. Do we really want BMD to spend resources chasing the lowest common denominator?

I fully acknowledge that the BMD team has ton of requests flying at them. And they must have their hands full prioritizing what needs to be done. And in actuality, the H.264 export out of Resolve is **way better** than what comes out of Avid Media Composer.

The problem though is defining what is good, better, best is a moving target when it comes to h.264/5, just see the comments about crf 18 vs 15. My main point is that I have never used an NLE (let alone DR) for final encoding. There are a lot of reasons, chief among them being crashes midway through. Of course, one has to export the timeline somehow, but that is what digital intermediates/masters are for especially when encoding for multiple formats like BD and the wide web world. The last thing I want to do is have to export a timeline with gpu intensive nodes more than once.

You don't have to stop at CRF=15. I used CRF=13, but my uploads were high end commercials which were fairly short, so had no problem with size or upload time. You can send ProRes master to youtube as well. Of course sending uncompressed is just waste of time.

I skimmed over everything and didn't see anyone mention rendering out of resolve in h.264 UHD, 2 pass, high profile etc? why wouldnt you just do that straight out of resolve?

That would be a very good reason for the next version of Resolve to have the option to export directly to ffmpeg/ffmbc.

I'd be happy if they just gave us more control over the encoding process, like CRF mode, and the various flags that can be set.

just to add a little more information to the conversation take a look at how much de-noising can effect the file size.

https://mattgadient.com/2013/06/29/in-d ... -examples/

Personally I think this is just one of many things that points to the fact that there are no universal best settings.

Noise is a well known issue for encoding algos. And there are many more for the unwashed masses. Like grading it is a bespoke process and professional encoders mastering a BR will often tweak the settings for each individual scene in a movie vs relying on a simple 2-pass algo. Of course, the studios have the resources for that. Crf is a good proxy but you may still end up with a dodgy scene or transition. The ability to tweak the settings for those parts is key. Not possible from within an NLE like DR.

just to add a little more information to the conversation take a look at how much de-noising can effect the file size.

https://mattgadient.com/2013/06/29/in-d ... -examples/

Personally I think this is just one of many things that points to the fact that there are no universal best settings.

Exactly, that why I said suggesting single bitrate value is not "perfect". Suggesting CRF value is much better, but you may end up with big (to big) files as side effect.
As you said- it's all about master nature and mainly about noise. My mentioned case for 70Mbit was exactly such a case- advert which had added fake heavy grain to simulate film look. Many others (clean ones) produced e.g. 20-30Mbit average bitrate, this one "caused" 70mbit, but all grain was kept very well. This is what I call transparent encode. Youtube of course killed a lot of this grain, but I always want to give it "good" master, rather then already over compressed one. CRF mode is a good solution for providing it, without worrying to much and using custom setting for each source.

Noise is a well known issue for encoding algos. And there are many more for the unwashed masses. Like grading it is a bespoke process and professional encoders mastering a BR will often tweak the settings for each individual scene in a movie vs relying on a simple 2-pass algo. Of course, the studios have the resources for that. Crf is a good proxy but you may still end up with a dodgy scene or transition. The ability to tweak the settings for those parts is key. Not possible from within an NLE like DR.

CRF won't really break single scene as it can change bitrate from cut to cut even 3x. At <15 values even transitions, fades etc (which are rather week point of x264) should look very good. If not good enough you use CRF<10 and this will be as good as eg. ProRes (of course your bitrate will jump to e.g. 100Mbit+ for HD).
Need for use 2 pass mode+ additional human assisted passes is purely based on fact that BD is still limited to bitrate which can't always delivered expected quality. There are masters when even close to 40Mbit is not enough and these are the ones which you need to tweak manually. I can tell you that probably 80% BDs (specially from big studios) will be done just with 2 pass encode without any manual tweak. Specially today as no one wants to spend time and money for this final touches. To be fair many modern masters (specially clean ones) don't really need any special touch if good BD encoder is used and average bitrate is above 30mbit.

CRF won't really break single scene as it can change bitrate from cut to cut even 3x. At <15 values even transitions, fades etc (which are rather week point of x264) should look very good. If not good enough you use CRF<10 and this will be as good as eg. ProRes (of course your bitrate will jump to e.g. 100Mbit+ for HD).
N

This is why i switched to h.265 as soon as you tube started accepting it. I can get the same quality and a smaller file, or better quality and the same sized file as h.264.

Yes, but encoding time is also much bigger.
At certain bitrates h265 actually make almost no sense, specially against x264.
h265 is good for low bitrates, when you really trying to squeeze a lot into relatively tiny file.
For "intermediate" quality I think x264 is still better choice (quality v. speed).
Did not test it recently, but x265 use to create semi good results at high (e.g. 30Mbit+ for HD) bitrates compared to x264, specially for grainy/noisy sources.

Yes, but encoding time is also much bigger.

What where you testing with?

Using ffmpeg and tapping into the nvidea api, I can trans-code UHD DNxHR HQ to h.265 at about 60fps. Interestingly my gtx 1070 is the bottleneck.

x265 not Nvidia. If you use Nvidia h265 with higher bitrates I doubt it will be any better than x264.

CRF won't really break single scene as it can change bitrate from cut to cut even 3x. At <15 values even transitions, fades etc (which are rather week point of x264) should look very good. If not good enough you use CRF<10 and this will be as good as eg. ProRes (of course your bitrate will jump to e.g. 100Mbit+ for HD).
Need for use 2 pass mode+ additional human assisted passes is purely based on fact that BD is still limited to bitrate which can't always delivered expected quality. There are masters when even close to 40Mbit is not enough and these are the ones which you need to tweak manually. I can tell you that probably 80% BDs (specially from big studios) will be done just with 2 pass encode without any manual tweak. Specially today as no one wants to spend time and money for this final touches. To be fair many modern masters (specially clean ones) don't really need any special touch if good BD encoder is used and average bitrate is above 30mbit.

I am not sure how what I said differs from your response? If I have a specific scene or transition that needs a CRF<15, but the rest is fine at 15 or greater, how does one achieve that in an NLE? And if bitrate was the only consideration, I would just encode everything for BD at the max CBR. It's a little like having a hammer and thinking everything is a nail.

Ok guys, before anyone says anything about YouTube compression and loss of quality, I just wanna say that I have seen plenty of videos on YouTube that are very sharp and crisp.
So now, I am using a Mac? I rendered out a 4k/24fps video which looks very sharp and crisp on the timeline. However, once it's rendered out using YouTube settings, the quality is NOT as sharp and crisp. And after uploading to YouTube, there is some loss of quality. I tried different settings other users had posted here but they are not as sharp or crisp after render.
How are these videos on YouTube getting these sharp and crisp quality?
Please see the image below as it shows my video and audio render settings for YouTube.

What render settings should I use on a Mac for YouTube?
On Render page under Video tab what settings should I use?
Format: QuickTime or MP4?
Codec: H.264 or something else?
Quality: 320 Kb/s or higher?

Under Audio Tab:
Codec: AAC or Linear PCM?
Data Rate: 320 Kb/s or higher?
Bit Depth: 16 or higher?

Render Speed: Should it be set to Maximum or 50?

Thank you all so much in advance,

ZIA

Ok guys, before anyone says anything about YouTube compression and loss of quality, I just wanna say that I have seen plenty of videos on YouTube that are very sharp and crisp.
So now, I am using a Mac? I rendered out a 4k/24fps video which looks very sharp and crisp on the timeline. However, once it's rendered out using YouTube settings, the quality is NOT as sharp and crisp. And after uploading to YouTube, there is some loss of quality. I tried different settings other users had posted here but they are not as sharp or crisp after render.
How are these videos on YouTube getting these sharp and crisp quality?
Please see the image below as it shows my video and audio render settings for YouTube.

What render settings should I use on a Mac for YouTube?
On Render page under Video tab what settings should I use?
Format: QuickTime or MP4?
Codec: H.264 or something else?
Quality: 320 Kb/s or higher?

Under Audio Tab:
Codec: AAC or Linear PCM?
Data Rate: 320 Kb/s or higher?
Bit Depth: 16 or higher?

Render Speed: Should it be set to Maximum or 50?

Thank you all so much in advance,

ZIA

If you want a magic bullet setting that is going to look good no matter what...

1. Export out of Resolve in ProRes 422 (HQ)
2. Use FFMPEG and encode at a CRF of <15.

What the comments above were saying is that YouTube will recompress everything. What you want to avoid is having YouTube recompress an already heavily compressed file.

In an ideal world, you would simply unload the ProRes file to YouTube, then it only goes through heavy compression once. Unfortunately, this isn't realistic lots of the time due to lack of bandwidth, so the next best thing is to give YouTube a slightly compressed file, the FFMPEG compressed method mentioned above.

The screen shot for the video resolution shows HD rather than 4K so that may be related to lack of 4K sharpness?

Ok guys, before anyone says anything about YouTube compression and loss of quality, I just wanna say that I have seen plenty of videos on YouTube that are very sharp and crisp.
So now, I am using a Mac? I rendered out a 4k/24fps video which looks very sharp and crisp on the timeline. However, once it's rendered out using YouTube settings, the quality is NOT as sharp and crisp. And after uploading to YouTube, there is some loss of quality. I tried different settings other users had posted here but they are not as sharp or crisp after render.
How are these videos on YouTube getting these sharp and crisp quality?
Please see the image below as it shows my video and audio render settings for YouTube.

What render settings should I use on a Mac for YouTube?
On Render page under Video tab what settings should I use?
Format: QuickTime or MP4?
Codec: H.264 or something else?
Quality: 320 Kb/s or higher?

Under Audio Tab:
Codec: AAC or Linear PCM?
Data Rate: 320 Kb/s or higher?
Bit Depth: 16 or higher?

Render Speed: Should it be set to Maximum or 50?

Thank you all so much in advance,

ZIA

If you want a magic bullet setting that is going to look good no matter what...

1. Export out of Resolve in ProRes 422 (HQ)
2. Use FFMPEG and encode at a CRF of <15.

What the comments above were saying is that YouTube will recompress everything. What you want to avoid is having YouTube recompress an already heavily compressed file.

In an ideal world, you would simply unload the ProRes file to YouTube, then it only goes through heavy compression once. Unfortunately, this isn't realistic lots of the time due to lack of bandwidth, so the next best thing is to give YouTube a slightly compressed file, the FFMPEG compressed method mentioned above.

Maybe you can tell me if I have been going about my system all wrong. My videos I upload onto youtube are genereally shot in 4k/24fps H.264 mp4 with 100mbps bitrate- these are my cameras settings when filming. I will render out of resolve at 4k, H.264 mp4 and upload onto youtube. My videos are generally 2-5 minutes long. I've only been oding this for about two years, and to my eye the videos look pretty good on youtube. I have abviously seen better on there, and I've also seen worse. Am I sacrificing quality with my method? What can I do to improve the way I do it? Would the above advice fit my situation as well?I appreciate any help. Thanks.

I've only been oding this for about two years, and to my eye the videos look pretty good on youtube. I have abviously seen better on there, and I've also seen worse. Am I sacrificing quality with my method? What can I do to improve the way I do it? Would the above advice fit my situation as well?I appreciate any help. Thanks.

Hey Joshua
There are many things that can improve quality, from how you light, the lenses you use to what you shoot, how you grade it and how you eventually export it.
Since I didn't see your work its hard to give any specific advice... so here's a general one

I'm assuming you're shooting on some sort of mirrorless/DSLR (which one?), and most probably 4:2:0. One thing you can do is start using some sort of external recorder and shoot Prores instead of 264. Even if most small cameras (not all) truncate the HDMI output to 8 bit, you'd still get better compression, higher bitrates and most importantly - 4:2:2. Starting off at a higher quality goes a long way... and ext' recorders aren't that expensive .

And if you shoot with any 8 bit camera, don't expect too much from grading.
Try to get as close to your intended look as possible by careful lighting and, if available, camera profiles.

And if you shoot with any 8 bit camera, don't expect too much from grading.
Try to get as close to your intended look as possible by careful lighting and, if available, camera profiles.

True.. though from past experience with better 8 bit formats (XDCAM, DVCPRO or even DNX in lower variants), 4:2:2 and better compressions go a long way too. especially when it has to do with multi-generational decay

Absolutely. I was rather thinking about 8 bit and heavy compression.
Even Sony's XAVC-S is a significant improvement over AVCHD, but I wouldn't use it with Slog3, which is rather a marketing gimmick than a serious option.

I've only been oding this for about two years, and to my eye the videos look pretty good on youtube. I have abviously seen better on there, and I've also seen worse. Am I sacrificing quality with my method? What can I do to improve the way I do it? Would the above advice fit my situation as well?I appreciate any help. Thanks.

Hey Joshua
There are many things that can improve quality, from how you light, the lenses you use to what you shoot, how you grade it and how you eventually export it.
Since I didn't see your work its hard to give any specific advice... so here's a general one

I'm assuming you're shooting on some sort of mirrorless/DSLR (which one?), and most probably 4:2:0. One thing you can do is start using some sort of external recorder and shoot Prores instead of 264. Even if most small cameras (not all) truncate the HDMI output to 8 bit, you'd still get better compression, higher bitrates and most importantly - 4:2:2. Starting off at a higher quality goes a long way... and ext' recorders aren't that expensive .

Thanks for the tips. I primarily do videos for my real estate business as well as do some for some colleagues. I also really enjoy making little "shorts" of areas where I live as a hobby. The two pieces of equipment I use is a DJI Phantom 4 Pro and I use a DJI OSMO pro with a micro 4/3 15mm f1.7 lens, other lenses available. The encoder options out of camera are MPEG4/AVC/H.264, and formats available MP4/MOV. I definetly want to be able to take my craft to the next level and have potential to grow. I know from a quality standlpoint I can get better, but for what I do having portability flexibility and having a stabilised gimbal is a must. When I purchased the OSMO i was considering getting a GH5, but then I would have needed a system to stabilize it, plus I liked the ability to be able to use the Zenmuse X5 camera on multiple platforms in the future. I'm open to getting an external recorder, do you have suggestions for that. Knowing the applications I use it for I would like something portable, and it doesn't need a monitor on it as I already have a monitor with the system I have, the DJI Crystalsky, which have HDMI outputs that a recorder could be connected to. I know there are a ton of high dollar ones out there, but I don't even know where to start or what the are officially called in order to search for one. I would like to stay under the 1K price range, obviously the least expensive while still retaining good quality is best, but I do know that you do get what you pay for. Any help would be appreciated, and thanks for the help so far.