ShadowPlay -> Youtube Video Quality Tests

[Fox Parker]

As most of you know, I've been toying with Nvidia's ShadowPlay and when directly uploaded to YouTube, there are compression artifacts in 1080p that just shouldn't be there, and whether any of the lower resolutions even work is a complete gamble.

I've taken a match on River City Night (dark scenes and monochrome scenes tend to be hard on compression algorithms) and uploaded three copies to YouTube. The first copy is straight from ShadowPlay. Second and third copies are an attempt to translate YouTube's suggested settings for other programs to Handbrake. The second video was encoded with those settings and at 60 frames per second. The third and final video was encoded with those settings and at the 30 frames per second recommended by YouTube.

What I want you guys to do is compare each of the three videos at various resolutions, but especially 1080p. Let me know if you can identify any specific flaws, and which you think came out the best.

Direct From ShadowPlay:


Suggested Settings at 60FPS transcoded by Handbrake:


Suggested Settings at 30FPS transcoded by Handbrake:

[devilcrayon]

Other than the hang in playback near the beginning when you come down out of base, the original otherwise looks best.
the A version definitely looks lower resolution more poorly smoothed: text and icons have a soft edge they shouldn't.
the B version has the same softness, plus stuttery motion-blurred movement on the Highlander.

[Fox Parker]

I see all sorts of compression artifacts in the original. *Especially* around text. Watch the Awesome at the end, and pay close attention to the target designation text. You'll see it become unreadable at times. Similar issues with the UI. All of the things I've mentioned are lessened or gone in the later two tests. Are you watching it in 1080p?

Further, I've done several straight-from-ShadowPlay uploads and its been a gamble whether any of the resolutions below 1080p are even watchable. 480p on one upload is completely garbled. So either way I have to transcode them before uploading. Another note is that YouTube complains about the ShadowPlay file. And finally, all three of the above videos look crystal clear when played locally. Its YouTube making things difficult.

[JD McFrizzle]

i thought A looked far sharper than the original. didnt notice any "artifacts", just seemed clearer to me. could be my low graphics settings, but i thought the original was the worst.

[cyberstrabadi]

In the other hand the original looked better to me but the difference was less than minimum.

[Fox Parker]

I've loaded each copy and stopped them at exactly the same point. I then took screen captures of each. I loaded them all in the same layered file, one version per layer. I then added a layer with a screen capture of the original ShadowPlay file.

Based on the location I captured, all of the YouTube videos are nearly identical. The only difference in this particular capture is the exact 'rotation' of the blurry spots. The capture was done at timestamp 1:41, game time remaining 14:40, and distance to targeted location 388m.

I also checked the section with the highest amount of compression artifacts, when finishing off the two Jesters. All three YouTube copies are extremely blurry.

The conclusion is that there is no real appreciable difference between them all for this night video. I also came to the conclusion that major YouTube personalities have some sort of quality trick not available to me. The clarity in my source file played locally is a hundred times better than after YouTube transcodes it. Which sucks because I've watched plenty of videos on YouTube in 1080p that were as clear as my source file. Not too happy there.

[Javenri]

Try www.dailymotion.com to see any differences in encoding. Signing up there is a breeze compared to youtube.

[JD McFrizzle]

that daily motion vid you posted jav had me buffering every 2 -5 seconds unless i put it at lowest quality. Ive never had those "buffering breaks" from you tube even on 1080p quality..dunno why.

[Fox Parker]

Here are the first set of stills that I used for comparisons. I made other stills but didn't save those in the end.
Be sure to view them full screen or download them and make them layers in a new image file and switch between the layers.

ShadowPlay source file on Hard Disk

YouTube upload straight from the original ShadowPlay source.

YouTube upload A from locally transcoded file using YouTube suggested settings.

YouTube upload B from locally transcoded file using YouTube suggested settings.

[cyberstrabadi]

NVIDIA has a new update yesterday that in description among other things say that they fixed some shadowplay stuff including the fps glinch.Have you downloaded it?It could be fixing your issues.