Video compression

Good morning... when I upload a video of +/- one Gb in mp4 - h264, to youtube, for example, and then I download it from there, this video with the same specifications can have a much smaller size but visually with the same quality as the one I uploaded.

When I try to compress the original 1gb to a similar size to the one downloaded from youtube I only get very poor quality results.

Does anyone here know what is the "recipe" youtube uses to compress a lot with optimal quality?

Any recomendation about?

Very warm regards

Federico

Good morning... when I upload a video of +/- one Gb in mp4 - h264, to youtube, for example, and then I download it from there, this video with the same specifications can have a much smaller size but visually with the same quality as the one I uploaded.

I'm neither a YouTube nor codec expert, but I believe that YouTube uses a more efficient codec than H.264. You could try exporting your Resolve project using H.265, for example, which is quite a lot more efficient than H.264.

If you uploaded in higher resolution than HD, YT is using a better codec.
But from my experience, it doesn’t look better than what I can achieve with x.265 (HandBrake) at the same bitrate.

Does anyone here know what is the "recipe" youtube uses to compress a lot with optimal quality?

h265, 4k, 60mbit for 24/30fps

https://www.reddit.com/r/videography/co ... &context=3

Does anyone here know what is the "recipe" youtube uses to compress a lot with optimal quality?

Youtube doesn't really care about optimal quality. During Covid they lowered the quality even more and most people didn't even notice.

If your files look worse I'd worry more about that.

Second that.
(And please don't shout – all caps – your post is not more important than everybody's)

Youtube doesn't really care about optimal quality. During Covid they lowered the quality even more and most people didn't even notice.

And, interestingly/cynically they're now rolling out 1080p-Premium playback resolution which is only available to those with a YouTube Premium subscription. To create that they simply slashed the bitrate on regular 1080p streams -- which is probably the downgrade you noticed.

I don't know about that but during Covid to stop the internet melting with everybody working from home they lowered the quality. I guess that's two years ago now. Three? Yikes.

Thank you all for the useful information you gave me.

fp

You'll generally bet much better results using something like Voukoder's h.264 than the built-in one Resolve has.
I get just as good results with much smaller sizes.

Second that, Voukoder is based on ffmpeg, just like HandBrake, but integrated.
I wish it existed for MacOS.

It is my understanding YT uses VP09. Also, I've heard you will get better quality when uploading in AV1.

It is my understanding YT uses VP09. Also, I've heard you will get better quality when uploading in AV1.

I found no significant change,

Screenshot 2023-06-15 160455.png (61.01 KiB) Viewed 2506 times

I have a detailed write up, but its on reddit, so not available any more. But basically I uploaded equivalent bitrates in H.265 and AV1 to YT, Downloaded the files and performed VMAF analysis of then vs what I sent

If you have a fast internet connection, just upload DNxHR HQ. No downside to AV1 if you have hardware acceleration for it. But you still need to upload 60mbit+ 4k 24fps files, so you won't save any upload speed.

It is my understanding YT uses VP09. Also, I've heard you will get better quality when uploading in AV1.

I found no significant change ...

Thank you for sharing that If memory serves me well I've watched one YT video in which he was showing clear difference with his high detailed gaming footage. Could it be that your result was due to footage you used?

If you have a fast internet connection, just upload DNxHR HQ.

What was your experience / quality differences you found when using DNxHR HQ, please?

Thank you for sharing that If memory serves me well I've watched one YT video in which he was showing clear difference with his high detailed gaming footage. Could it be that your result was due to footage you used?

This is the footage, was intended to be challenging to compress.

What was your experience / quality differences you found when using DNxHR HQ, please?

The DNxHR file was VMAF score of 95.3 vs 92.3 for 84mbit AV1. I doubt a human could see this difference.

I choose to upload 60mbit h.265 encoded with Nvenc.

Thank you for sharing that If memory serves me well I've watched one YT video in which he was showing clear difference with his high detailed gaming footage. Could it be that your result was due to footage you used?

This is the footage, was intended to be challenging to compress.

What was your experience / quality differences you found when using DNxHR HQ, please?

The DNxHR file was VMAF score of 95.3 vs 92.3 for 84mbit AV1. I doubt a human could see this difference.

I choose to upload 60mbit h.265 encoded with Nvenc.

Thank you again!

Hi, that's a good point about video compression at YouTube. I noticed that they mainly change the resolution after the video upload, and I cannot notice actual compression when comparing it to my original footage. For my personal projects, I use online video editors—it saves time. You can try this Videmix.com website I recently found for video compression.