Does anyone have any suggested settings in order to get non pixely video from the Action when uploading to YouTube?

I've been messing around with a bunch of export variations and all of my videos look a little pixellated.  

Playback on my machine looks great, but not on YouTube.

I want to use 1080p not 4K when uploading.

As an example...   

Thanks,
Andrew

Hi Andrew.
No matter what you do, the resulting file on YouTube will never look as good as the file that resides on your computer. YouTube will recompress the video on upload to reduce server space and for rebroadcast bandwidth and there will be some degradation of the resulting file as it's a copy of a copy. If your editing program permits it, try choosing a 4K resolution output for export but reduce the video proportions down to 1080 x 1920. This will increase the minimum, maximum and overall bit rate of the file as will selecting 'Best' quality for export but the resulting file size will be huge. The upload should then not suffer as much from Youtube's compression and may look a bit better, but still not the same as you see looking at the original file. Using HEVC for export seems to give me the best results for YouTube but that may vary from editing program to editing program.

I have been having the same problem.  I've not found any really good solution.

But I also wanted to comment on the video - very nice.  I still find the RockSteady quite impressive.  On the dirt track and over the plank bridge - nary a bounce or shake. If you don't mind my asking - what bike mount did you use?  I'm still new to action cameras and don't have any accessories.

Solution is simple - you have to export your video file to minimum 1440p (2.7k). Than if you upload video to YouTube you'll force to use vp9 codec instead of avc which is worse. I've check nerd stats and your YouTube video use avc codec and that's the reason of pixelated video. I've done some test and it works !!! Check my video (it's not recorded by Osmo action)
youtu.be/qsmneF05Kaw  - 26-30 sec. Look at the grass - so pixelated.
youtu.be/IATwA1rsc4Q - the same video but uses vp9 codec.
That's all

youtube recode all video to their own format, if you have a very popular channel this will not happen or at least to a lesser degree so you footage look better.
For a wile i have edited 1080p footage to 4K HEVC as youtube don't  rape that as much.

Of course if you are on a internet connection with moderate speed it is a pain some times, i myself only have 60 mbit upload, but a whopping 1 gbit download.
My next internet connection will be 1/1 or 5/5 gbit on fiber.

peiter Posted at 6-4 00:43
youtube recode all video to their own format, if you have a very popular channel this will not happen or at least to a lesser degree so you footage look better.
For a wile i have edited 1080p footage to 4K HEVC as youtube don't  rape that as much.

Hi Peiter.

I don't know where you got that info about popular channels not being compressed or being compressed less but that is wrong.

Every video uploaded to YT gets compressed and every individual resolution gets compressed the same for all YouTubers.

Cheers,
Dave.

Just something i have heard here and there, and IMO it looked valid as i myself have noticed channel content that appear to be much better than other, and which i cant just say thats due to the camera and bitrate / compression formats on the original footage.
but comparing it have of course never been with the exactly same footage, so there are that unknown factor.
It may be wrong in the end, but really google i trust very little
If i knew someone with a popular channel i would love to do the test and UL the same footage to that channel and my own, and then see what run best.
Of course i would never go to youtube for quality video, just like i would never expect that from one of the streaming companies ( which i only know of from others cuz of course i would not pay money for such garbage video quality and content )
Youtube videos "often" freeze here at least 4K, and they have the audacity to suggest maybe my internet connection are not so good, and i am on a 1 gbit connection that will move around 125 megabytes every second, and my next connection will probably be 5/5 gbit and youtube will probably still by then say you might have some connection problems,,,,,, which i have never had on fiber or like now on docsis 3.1 cable.

All videos are compressed but some uses avc and some vp9. And he is right - popular youtubers movies are always coded with vp9. It gives you better quality than avc. I've watched some examples on YouTube, and I've done my own test - I've exported 1080p to 1440p (2.7k), than uploaded to youtube and video quality is better because of vp9. Check my samples and believe.

Hi.

Here's a few points that may help.

YT will compress everything that you upload, there are no exceptions.

HD and below will be h.264, above will be VP9

Both YouTube's h.264 and VP9 are inter-frame codecs and will both suffer similar issues.

Although you won't get better quality from VP9 it will look subjectively better than h.264 at lower bitrates at the same resolutions due to it being more efficient, similar to h.264 VS h.265

Because you are using destructive inter-frame, lossy, compression with h.264 with your Action files and so is YouTube with its output, you need to feed YouTube an intermediate master that doesn't degrade between you and YouTube.

For the least amount of encoding artefacts in your final YT output you should use an intra-frame codec as your intermediate master. This will reduce artefacts but will not get rid of them completely.

You can also use an intra codec that uses lossless compression, you don't have to use uncompressed. Things such as HQX, HQ, DNX, ProRes etc. at the appropriate settings.

These files will all be very large compared to the original files.

If you've only shot HD, stay HD. Upressing to 4K to fool YT is a bit of a myth, you'll likely loose more information in the uprez compared to what you'll gain at the YT end. If you know enough about upacaling and what algorithms to use for specific motion characteristics, maybe give it a try as you may notice a slight difference in heavily compressed detailed blocks, but highly unlikely.

In the case of Osmo Action footage you'll never see any similar footage to yours looking any better if you flollow my intermediate suggestion. All the problems are in the original footage. These are problems due to the compression method and not the camera, these problems simply get worse with subsequent lossy encodes.

If you ever see footage similar to your own with an Osmo Action that looks better in areas that break up, macro-blocking, and you've followed my upload workflow, it's likely that the master has had a slight blur applied to help with the YT encode.

If you ever see amazing looking footage on YouTube, not actioncams etc. This is likely to be professional cameras shooting uncompressed, YUV, RAW, or a lossless codec and then uploaded from a similar intermediate master.

Bottom line. The vast majority of footage shot is using inter-frame compression, phones, actioncams etc. They all have issues to a certain degree and all need treating in a certain way to get the best out of them. Unfortunately there's no short cuts and huge file uploads are always going to be needed.

The only way around this is to either just upload the Osmo Action files 'as is' which will be the best quality upload. Or use an NLE that smart renders back to the original codec parameters for cuts only editing.

Hope this helps.

Cheers,
Dave.

David you're right. I know that upscaling will not give me better image quality. I used it to force YouTube vp9 which is more efficient - like you said. And please watch my samples and look for 26-30 sec where grass is looking like a mess (avc) and way batter with vp9 codec. Which one looks better?

djiuser_iSwZn377A1Jz Posted at 6-4 02:46
David you're right. I know that upscaling will not give me better image quality. I used it to force YouTube vp9 which is more efficient - like you said. And please watch my samples and look for 26-30 sec where grass is looking like a mess (avc) and way batter with vp9 codec. Which one looks better?

Everything is up to a point subjective. A lot will not also depend on the original file upload quality and subsequent compression by whichever cloud service you upload to but also which browser is used for playback, monitor and graphics capability of the viewer.
Take for instance your example videos.  You may well be seeing a difference and I won't dispute that but Iat my end I ran both videos in sync on two calibrated monitors, side by side and running the 2.7K alongside your 1080p examples I would have said they looked roughly the same to me.

Wheres upon looking at your 2.7K video at 1080p made that version look considerably worse on my screens.

I know at this point that David may not feel this is a good comparison as screen grabs seldom are, and then placing them on this site doesn't do the resulting images any favors either. This is why I personally feel upscaling has limitations, but allowing the code the same average overall bitrates while constraining the file to its original screen size can lead to less information being lost of the file when it's encoded at export. But, and this is a very strong but, how much additional information you can encode compared to the additional visual acuity gained for the additional file size needed may be questionable.
Just to add that the screen grabs took place at 20 seconds and not the suggested 26-30 seconds you suggested.

Well, static screens shows that not upscaled 1080p (avc) is a bit sharper, but when its moving it looks like minecraft sometimes (blocks everywhere) that ruins video. Ok Andrew, its up to you - just give it a try, and maybe you could upload some samples with upscaled video here? Than everyone could compare and choose.

djiuser_iSwZn377A1Jz Posted at 6-4 02:46
David you're right. I know that upscaling will not give me better image quality. I used it to force YouTube vp9 which is more efficient - like you said. And please watch my samples and look for 26-30 sec where grass is looking like a mess (avc) and way batter with vp9 codec. Which one looks better?

Hi.

I've had a look and they both look the same at 1080. Neither one at 1080 is VP9 either.

This is what I was getting at before, there is no need to upscale 1080 to anything higher for a 1080 target, it won't make a difference.

Anything above 1080 from a 1080 upscaled to 4K or 1440, will simply have more bitrate applied to it, which will help with certain encoding artefacts. But like I said,  there will be no difference to the 1080 version and most people have problems streaming 1080 and won't be streaming higher just to get a higher bitrate version of an upscaled 1080 version.

Cheers,
Dave.

BTW.

For anyone who thinks that bigger YouTubers get preferential treatment with their videos with regard bit rate and quality, this isn't the case.

Bigger YouTubers don't get VP9 encodes of low resolutions just because they are bigger YouTubers. VP9 is used for certain resolutions and it's also impossible to use it on certain platforms and apps, as they simply can't decode VP9. It's similar to H.265 in that sense, it's harder to decode compared to h.264 and more so with older hardware.

I've just downloaded the latest video by the biggest YouTuber and the low res ones were h.264

Cheers,
Dave.  

Whatever. I give up.

avc 1080p


vp9 1440p

Do you also have it with footage from other camera's?

djiuser_iSwZn377A1Jz Posted at 6-4 05:26
[view_image]
Whatever. I give up.

Hi.

I don't think anyone is questioning what you're seeing, at least I wasn't.

VP9 will get used for certain resolutions and frame rates, but not all the time.

You can watch the same video on different devices and platforms and you may be seeing a different codec even at the same resolution etc. I've run similar tests to what you are showing and have seen it change from device to device.

Unfortunately, YouTube themselves chop and change the outputs, so it's not always easy to say exactly what they're doing.

For instance, I looked at your video and three different things said mp4, although I think my TV would see it as VP9, even at 1080/60.

The one thing that is constant though is that the 1080 version will always suffer from low bitrate and an upscaled version will have more bit rate that wont gain in resolution but may help reduce the effects of macro blocking due to higher bit rate. But like I already said, the vast majority of YouTube streams are less than 1080 and most people can't or wont increase the resolution above 1080, even if there is a genuine 1:1 UHD/4K version.

You've also got to remember that outside of some geeky types, the huge majority of viewers can neither see the difference, no matter how obvious it is, or simply don't care. Especially those watching cat videos.

Cheers,
Dave.

djiuser_iSwZn377A1Jz Posted at 6-4 05:35
avc 1080p
[view_image]

Here's something you may find interesting, although it does raise more questions.
Cheers,
Dave.

David_Harry Posted at 6-4 07:32
Here's something you may find interesting, although it does raise more questions.
Cheers,
Dave.

Hi David.
Yes, I found it interesting and made me query my own advice, so I made an export or one of my own videos as I had recommended earlier to the point where I could see a distinct improvement on my PC. Clothing had more detail, the grass looked sharper, etc. I uploaded the result to YouTube and... absolutely no visual difference to the earlier one I had put up. No difference to the codec in use either. There seemed to be a difference in the bit rate of the playback data stream. It was slightly larger, but with no visual improvement that I could see. I will try upscaling later but I'm getting this sinking feeling that that won't actually work either as far as trying to improve the visual data.
So please ignore my earlier advice. It worked for Flickr but not YouTube.

Ray-CubeAce Posted at 6-4 08:46
Hi David.
Yes, I found it interesting and made me query my own advice, so I made an export or one of my own videos as I had recommended earlier to the point where I could see a distinct improvement on my PC. Clothing had more detail, the grass looked sharper, etc. I uploaded the result to YouTube and... absolutely no visual difference to the earlier one I had put up. No difference to the codec in use either. There seemed to be a difference in the bit rate of the playback data stream. It was slightly larger, but with no visual improvement that I could see. I will try upscaling later but I'm getting this sinking feeling that that won't actually work either as far as trying to improve the visual data.
So please ignore my earlier advice. It worked for Flickr but not YouTube.

Hi Ray.

What NLE are you using?

Cheers,
Dave.

David_Harry Posted at 6-4 09:05
Hi Ray.

What NLE are you using?

MEP Premium 2019 David.

David_Harry Posted at 6-4 09:05
Hi Ray.

What NLE are you using?

As an aside David.
Looking at both files I produced using the same HEVC codec. The first one using a higher set of bit rates but keeping the video size to 1920x 1080 MediaInfo reports the Codec ID as being isom (isom/avc1) but the upscaled version still using HEVC reports the Codec as being iso4 (iso4/hvc1).
Would that header make a difference as to how YouTube handles the file for compression?
The file itself is not that impressive when viewed on my PC But I will upload it to YouTube now to see if YouTube uses the other codec.

Ray-CubeAce Posted at 6-4 11:34
As an aside David.
Looking at both files I produced using the same HEVC codec. The first one using a higher set of bit rates but keeping the video size to 1920x 1080 MediaInfo reports the Codec ID as being isom (isom/avc1) but the upscaled version still using HEVC reports the Codec as being iso4 (iso4/hvc1).
Would that header make a difference as to how YouTube handles the file for compression?

Hi Ray.

With a high enough bit rate you wont see any difference between h.264 and h.265 as long as you're using a high quality codec. For h.264 you can't beat X264

The problem is that no matter how good these encodes are you are still compounding inter-frame compression artefacts.

Using uncompressed or a lossless intra-frame codec is the only way to minimise such losses during subsequent transcodes/encodes.

But even when using an uncompressed/lossless/intra-frame codec as your intermediate YT upload file, you are still subject to the heavy compression of the final YT encodes. This will result in very quickly diminishing returns.

I use very large lossless uploads to get the absolute best but the differences can be very small. For me it's worth it but is usually very impractical for most people due to file size, data/bandwidth usage  and time.

I'm going to try and dig out an old video that may be useful.

Cheers,
Dave.

I have to admit, I am thoroughly confused.

As an average user, what should I set the Action's resol, framerate, etc.

After I get it to my computer and if I do NO processing, how should I upload to youtube?  

We will get into editing later! (g)

David_Harry Posted at 6-4 12:07
Hi Ray.

With a high enough bit rate you wont see any difference between h.264 and h.265 as long as you're using a high quality codec. For h.264 you can't beat X264

Hi Dave.
Yup, it's certainly a pita sometimes but I guess there are limits to all formats and what's cutting edge today will be mediocre in less than ten years. The problem is from a global perspective that most of the world is neither that well connected to the internet or can see the highest quality output. So, are we doing it for others or ourselves?

Ray-CubeAce Posted at 2019-6-4 15:53
Hi Dave.
Yup, it's certainly a pita sometimes but I guess there are limits to all formats and what's cutting edge today will be mediocre in less than ten years. The problem is from a global perspective that most of the world is neither that well connected to the internet or can see the highest quality output. So, are we doing it for others or ourselves?

Ok, so some of this is true, some not. I am a YouTuber with 551 subs. My friend is a YouTuber with over 100K subs. The more subs and views, YouTube will give you better quality. My friend's videos and my videos are shot in 1080/60 same settings and everything. His are VP9. Mine are the crappy one.

I tested it last night. I shot a 2.7K video and a 1080/60 video. I uploaded to YouTube using 4K settings with a bitrate of 40 mbps. They both came out in VP9.

Mamba334 Posted at 3-9 07:32
Ok, so some of this is true, some not. I am a YouTuber with 551 subs. My friend is a YouTuber with over 100K subs. The more subs and views, YouTube will give you better quality. My friend's videos and my videos are shot in 1080/60 same settings and everything. His are VP9. Mine are the crappy one.

I tested it last night. I shot a 2.7K video and a 1080/60 video. I uploaded to YouTube using 4K settings with a bitrate of 40 mbps. They both came out in VP9.

Check the video again after a few hours. Sometimes the processing continues for some hours after the initial upload before the highest quality appears. Again, without a good connection you won't get a preview at a higher bitrate than your connection allows but if you can see others 4K and 8K uploads then you should be able to see your own.  40mbps doesn't seem that high considering the output of the Osmo Pocket.  

Ray-CubeAce Posted at 3-9 08:37
Check the video again after a few hours. Sometimes the processing continues for some hours after the initial upload before the highest quality appears. Again, without a good connection you won't get a preview at a higher bitrate than your connection allows but if you can see others 4K and 8K uploads then you should be able to see your own.  40mbps doesn't seem that high considering the output of the Osmo Pocket.

youtu.be/QzSczAy3f9w

It worked for me

Ray-CubeAce Posted at 3-9 08:37
Check the video again after a few hours. Sometimes the processing continues for some hours after the initial upload before the highest quality appears. Again, without a good connection you won't get a preview at a higher bitrate than your connection allows but if you can see others 4K and 8K uploads then you should be able to see your own.  40mbps doesn't seem that high considering the output of the Osmo Pocket.

Well, his videos are over a few months old and shot in 1080/60 and showing VP9. When you select his little gear icon the highest is 1080/60. However, it looks just like a 2.7K video. He doesn't have that terrbile codec regular 1080/60 uploads have. I'll be uploading wqith 4K settings from now on. There's always a loophole lol.

Note, I read from another YouTuber that if the video is longer than 15 minutes, this trick won't work.

Mamba334 Posted at 3-9 09:20
Well, his videos are over a few months old and shot in 1080/60 and showing VP9. When you select his little gear icon the highest is 1080/60. However, it looks just like a 2.7K video. He doesn't have that terrbile codec regular 1080/60 uploads have. I'll be uploading wqith 4K settings from now on. There's always a loophole lol.

Note, I read from another YouTuber that if the video is longer than 15 minutes, this trick won't work.

The 15 minute rule could be true but I'm watching other YouTubers that run 4K live content with over an hour show. Maybe because that have over a half million subscribers.  I think my longest video is 14 minutes.  I just checked one of my Osmo Pocket 1080p 50fps videos to upload and the bitrate at 1080p is an average of 81.4Mbps.
Here is the YouTube playback specs.

So yes, some further compression. Still looks almost the same as the original file though.

Ray.