FCPX 4K DOWNRES ISSUE

[Ronnie Amighetti]

Hi, I was playin' with my GX80 testing out a few different settings again to improve my workflow. I finally settled on: 

1. Shoot
2. Create a FCPX archive
3. Convert all media into ProRes422HQ resized to 1920x1080 with MpegStreamclip
4. Import all media and start edit

Anyway, while playin' with some ProRes for testing I noticed that FCPX is handling the resizing from 4K to 1080 in a bad way compared to the ProRes422HQ coming from MPEG Streamclip.

Please take a look at the frame grabs:

 

As you can see from the pictures, the ProRes version is quite a bit sharper. I put both clips straight into FCPX timeline.

Anyone?

BTW, I settled on 422HQ because it "only" increases file sizes by about 50% which is acceptable for my needs. Coming from a 4.2.0 camera @ 100mbps, resampling it to 4.2.2 with double the sampling rate (ProResHQ should be around 200somethingmbps from Apple specs) seemed logical. But I guess that's not the point with the "fault" I stumbled upon while testing...

[mercer]

Are you unclicking create optimized media on import?

[Ronnie Amighetti]

12 minutes ago, mercer said:

Are you unclicking create optimized media on import?

No optimized, no proxy, no nothing. Just the bare files, imported and straight to the timeline.

BTW I also tested with the 4K original against a ProRes422HQ 4K, in a 4K timeline, and there was no difference between the two. It only happens comparing the original 4K from the camera to the 1080p ProRes in a 1080p timeline. Hence why I was suspicious of FCPX downres algo.

To be honest, the frame grab I posted are some 400% magnification, but even looked at 100%, albeit there's a smaller difference in sharpness, the ProRes file just "pop" compared to the original.

[Axel]

1. The files transcoded by MSC match the timeline settings. They are played back as they are. Playing back HD ProRes (no matter the flavour) is the easiest thing for any Mac.

2. No matter if you chose Better Quality over Better Performance in the viewer, FCP X will always scale the quality down to give you realtime experience. It says better quality, not absolute quality, it's literally a preference of one over the other, not an exclusion. 

3. Therefore, it's normal that HD ProRes looks somewhat better than 4k mpeg4 *in preview*. Put the same clips side by side in your timeline, export them and then compare!

4. Mpeg Streamclip does not have it's own ProRes encoder. It's just an interface that let's you define the codec for export, but it uses QT, the same module that FCP X and Compressor use to transcode. No difference to triggering the process as background task from within FCP X. Only that this would have been more practical and smarter.

43 minutes ago, Ronnie Amighetti said:

BTW I also tested with the 4K original against a ProRes422HQ 4K, in a 4K timeline, and there was no difference between the two. It only happens comparing the original 4K from the camera to the 1080p ProRes in a 1080p timeline. Hence why I was suspicious of FCPX downres algo.

If they also play back equally smoothly in 4k, there wasn't any need to use optimized media anymore. Anyway, everybody's experiences with ProRes are that it's visually indistinguishable from the original. Just believe it and try what I suggest under point 3.

[Ronnie Amighetti]

Thanks Axel,

1. Yes, I agree

2. Yep, in fact I double checked to see if the viewer was Better Quality or Performance, it was the latter

3. Here comes the surprise: exported them and reimported and there is exactly the same difference in sharpness, hence the downscaling process also affects the export/output.

4. One clip plays smoothly, no problem whatsoever. A small project (so far I've only worked with musical videos) does not. That's why I'm looking for a more efficient way to handle the media.

Just a quick note: within Davinci Resolve there is no difference between the two clips. Same sharpness, same details.

Exported and reimported and they match.

[jonpais]

I saw one test where there was no discernible difference in quality between Pro Res HQ and Pro Res with the GH4 when outputting 4K to the BM Video Assist, and I don't have unlimited storage, so the latter would be preferable to me. I also thought that just dragging 4K clips to a 1080p timeline was not only faster, but the quality was also fine. 

[Ronnie Amighetti]

6 hours ago, jonpais said:

I saw one test where there was no discernible difference in quality between Pro Res HQ and Pro Res with the GH4 when outputting 4K to the BM Video Assist, and I don't have unlimited storage, so the latter would be preferable to me. I also thought that just dragging 4K clips to a 1080p timeline was not only faster, but the quality was also fine. 

Don't get me wrong, I always dragged 4K right into the timeline and couldn't notice anything strange. In fact if I had to give it a number I would say the quality difference is in the order of 1% or less. It just came to my attention because I was curious to see if there was any difference in working with the original files compared to the ProRes422HQ and that's what I found out. 

After reading around about the theory of downsampling 4.2.0 4K down to 1080 and getting sort of 4.2.2 I was curios to see if there was any gain in grading and color correction. That's when I discovered the sharpness difference between the two. I was looking for something completely different (banding and grading).

Coming from the audio world, I'm used to comparisons and tests so the first thing I did was putting the two clips on top of each other and set the composing to difference, just like I did thousands of times while testing audio plugins (we usually invert the polarity of one signal to listen for artifacts). That's where I noticed that colors where basically the same, but there was a noticeable difference in sharpness.

It just ticked my curiosity. That's all. Nothing major to be concerned I guess.

[Axel]

Very good thinking, Ronnie. I never tested this myself, but you're right: trust your own eyes.