NeatVideo - 'Blotchy' Output Problem

questions about practical use of Neat Video, examples of use
Post Reply
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

NeatVideo - 'Blotchy' Output Problem

Post by dpt »

Hi,

I've been amazed with how good the results with Neatvideo are since first buying the plugin, but I've recently noticed a quirk with the outputs that I'd like to resolve.

At times, in place of the original video noise, my video outputs appear to have an unwanted 'blotchy' effect.

I've made a short example video to show you exactly what I mean here - http://youtu.be/ldx0ipNnSAw

It's particularly noticeable in the middle-left of the frame, in what is the neutral background of the clip.

Worth knowing too, the effect appears to get stronger/worse as I increase the temporal radius (as demonstrated in the clip).

Have you come across this before? Is there a recommended strategy to avoid the blotchy effect whilst still eliminating noise?

Many thanks
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

Yes, I can see the blotches there. In the original (no Neat Video yet) version too.

These things are usually produced by the video codec used to compress the video. For example, when you apply a filter (any video filter, doesn't matter), the host application has to render the video (which includes applying the filter) and the compress it using a video codec to produce the output video file. That final compression is the main source of those blocky artifacts.

Apparently, the noise profile tells Neat Video that those blocky artifacts present in the original video (I can easily see them behind the noise in your youtube clip) are not noise and then Neat Video simply keeps them in. To avoid that I recommend to build a noise profile using a sample area that includes a good amount of those blocky artifacts. That may help to reduce them.

To avoid recreating new compression artifacts during final compression I recommend to try using different codecs and/or adjust the compression settings.

Hope this helps,
Vlad
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

Post by dpt »

Thanks for your reply.

In response to your suggestion via e-mail, I've made a new test video that includes a range of temporal radius' but also a test with Guassian Blur added to the original clip.

It's viewable here - http://youtu.be/AcKos0sUiYE

What do you extrapolate from this new clip please?

Some of this information is a tad tricky to get my head around I'm afraid, though I'm trying my best!

Cheers
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

The test with Gaussian Blur shows the same banding artifacts, which indicates that they are created by output compression (output codec). To see that better please increase the strength of Gaussian Blur (increase radius) and render out in two output formats:
1) using your regular codec;
2) using a lossless (perhaps even uncompressed) codec;
and then comprare the results.
I expect (1) to show strong banding and (2) show no such banding. If that is the case then there may be two solutions:
1) to use a better lossy codec that doesn't create such strong banding;
2) to use the same lossy codec but help it a bit by adding some artificial noise or by leaving some of the original noise in the video; yes, this may sound strange but that does help such a codec to avoid introducing compression artifacts, specifically banding.

Thank you,
Vlad
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

Post by dpt »

Thanks for your continued help.

I've turned up the Guassian Blur as high as possible and exported the same sequence in the following formats...

[1] Uncompressed 10-bit 4:2:2 - File Size 8.7 GB

[2] 1080p H.264 - File Size 45.5 MB

They both display the problematic banding to very similar extents.

What should we deduce from this please?
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

Which exactly version of FCP is it?

When you check preview in FCP itself, do you also see the same banding or is it only visible when you playback the resulting output clip in another player application (QT?)?

Thank you,
Vlad
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

Post by dpt »

I use FCPX 10.0.9.

Yes, the banding is visible when playing the rendered preview of the timeline back in FCPX.
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

So the same banding is visible in both FCPX preview and in the final clip played outside of FCPX, is that correct?

Is the banding visible in FCPX preview before you apply Gaussian Blur to the clip?

Vlad
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

Post by dpt »

Yes, the banding is visible in the preview of FCPX and in the exported files (both low quality and high quality exports).

The banding is not visible in the original clip, no. When I apply Gaussian Blur it becomes visible.

Of note, whilst the banding is highly visible during the playback of the video (both in the FCPX preview and with the exported files), it is not visible when the preview is paused in FCPX (however, it is visible in the paused exports).
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

When you check it in exports, which player software do you use to play back the clip? Is it QT? Sometimes QT adds its own banding to the playback, causing additional confusion.

Please try to re-import such clip back into FCPX and play it there. Will it also show the same banding or it will be less visible.

Generally, it looks like the banding shows up in FCPX when some (even built-in) filters are used. This may indicate a bug in FCPX, although I am not completely sure what causes the problem in this specific case.

Do you work with that clip in native format or you optimize it during import to FCPX? Perhaps that can make a difference.

Vlad
dpt
Posts: 8
Joined: Thu Mar 20, 2014 10:45 pm

Post by dpt »

I've tried using Preview (via Space Bar), QuickTime and VLC. The results appear to be the same across all three.

I've re-imported my two exports (hi-quality and low-quality) back into FCPX. Both have the banding visible when they're playing back in the timeline preview, but when I pause them both, the banding disappears in the hi-quality file.

The clips are imported into FCPX in their native format. The video was filmed using a Canon 60D DSLR.

I've just tried optimizing the native media (I highlighted the clips in the timeline, SHIFT+F to locate the original media in the event library, right-click and transcode the original clip to optimised media... I think that means I've done it? There doesn't seem to be any way of turning ON optimised media so I'm presuming it's used it automatically once the transcoding has occured).

I've then exported exactly the same timeline and found that it still has banding.
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

> but when I pause them both, the banding disappears in the hi-quality file.

That is very strange indeed. Looks like decoding of the video clip is done in different way during playback and during stopped preview.

In any case, the fact the problem of banding is reproducible with the built-in Gaussian Blur alone suggests that there may be a problem in FCPX, either in FCPX core render engine or in its interface that is used by effects like Gaussian Blur or Neat Video.

I think it makes sense to ask Apple support about this issue and perhaps even report it to them as a bug in FCPX. To do that from our side we would need a reproducible test case: a test project and clip allowing to demonstrate the problem. Could you send us (to support [at] neatvideo.com) a small sample original clip that is known to reproduce the banding with Gaussian Blur in FCPX?

Please also personally report this problem (based on the same clip and a test project with Gaussian Blur) to Apple using this page: http://www.apple.com/feedback/finalcutpro.html

If we both submit such requests to Apple there is a better chance that the problem is going to be fixed. Or perhaps they will suggest a workaround.

Thank you,
Vlad
Post Reply