Thanks Musicvid. I actually shot 4k for the ability to crop in and possible image stabilization. My 1080p Cineforms are always YUV 4:2:2 on a setting of High (not Filmscan). When shooting HD 1080p, I see no discernable difference from the originals, so I'll go with CCing the much easier to edit 1080p intermediates.

Final delivery will be high bitrate .MP4s for the web and same on a thumb drive for their HDTV.

Only thing I've read regarding possible color depth increase is from Barry Green's DVX200 handbook. However, it certainly makes sense that color resolution can't be increased if it isn't there in the first place.

My rig is fine for HD, but really struggles to push those 4k files around.

Thanks again for your help.

Thanks Musicvid. I actually shot 4k for the ability to crop in and possible image stabilization.

That is the best reason to do just what you are doing.

My 1080p Cineforms are always YUV 4:2:2 on a setting of High (not Filmscan). 

Cineform is a good intermediate codec, but some are better. If you are sold on 4:2:2 for space savings, UT 422 is a winner on the size/quality axis.

Do the UT 444 and 422 codecs allow for smooth playback like Cineform, but with smaller file sizes?

I do not think so.

CF is 'visual lossless' while UT is lossless (no data loss other than colour space conversion). In very rarely cases lossless can beat visual lossless in size (and only on synthetic data).

As for smooth preview playback -- your disks can easily be bottleneck because of intermediate files size. Also, UT is not the fastest lossless codec for 4K. Its decoding might also be limiting factor (you can check https://www.magicyuv.com/ for some decode speeds. Note that the numbers are for outdated version of UT and newer versions are a bit faster).

I'd stick to CF in your case.

Do the UT 444 and 422 codecs allow for smooth playback like Cineform, but with smaller file sizes.

If you meant smooth preview, the answer here is Yes. None of these codecs were designed for delivery. It is such an individual thing, I encourage you to run your own comparisons, and let us know what you find.

I would avoid UT 444 altogether, and use UT RGB for large, mathematically lossless intermediates that should preview quite well

Thank you both for your help.

Ken

@Musicvid, technically the OP isn't trying to get more bit depth out of their video, but reduced YUV compression. If you're sampling 4k colors at 1/4 their original resolution, that would be roughly 1080p, so you should notice less YUV compression when downsampling 4:2:0 4k to 4:4:4 1080p than you would from 1080p shot at 4:2:0, theoretically, correct? I wouldn't expect it to look as good as native 4:4:4 1080p, but definitely better than 4:2:0.

Unfortunately, interpolated pixels contain no more information than recorded pixels, except they are synthetic and there are fewer of them, and save for dithering loss, usually in the form of noise. Once again, theoretical gains are dwarfed by real-world inefficiency.

If the pixels are sampled at 1/4 resolution in certain color channels, then there shouldn't be any interpolation required if you are downsampling from 4x the resolution, righ

I should clarify, as reading the rumor mill online, I can see many people think you can get 10 bit 4:4:4 1080p out of 8 bit 4:2:0 4k, I do not believe this to be possible at all. I am strictly referring to 8 bit 4:2:0 4K to 8 bit 4:4:4 1080p.

This was discussed in various forums a few years back with introduction of UHD capable cameras like the Panasonic GH4, etc. I cannot remember all the details, but in the end the conclusion (for 4-2-0, 8 bit cameras) was shoot UHD and final render from UHD. For the OP, I would use 1080p lowest quality cineform proxies for editing and then switch them out to the UHD originals for final render. Anyway, works for me :)

The original idea was to see if there was any perceivable visual difference in the final .mp4 1080p deliverable.

I did a quick test: 1. CC 4k file, render to 1080p YUV 422 Cineform "fine" then to .mp4. 2. Render 4k to Cineform then CC that and render to final .mp4 (I put both Cineform files on the timeline and did a split screen render to the .mp4). I loaded the .mp4 test file onto a thumb drive and played it on a properly set up 42" HD TV.

There was no visual difference that I could see, even pixel peeping. The only 4k files I actually "corrected" on the UHD originals (due to pilot error) were a couple that needed stabilization and one that needed a slight rotation correction.

So, in this particular case, I see no need to spend the extra render time messing with the 4k files before making my 1080p intermediates. I can edit the full quality Cineforms without proxies.

If you use an intermediate your final output comes from two "generations" (ie orig. to intermed. to final)

If you use a proxy your final output comes from one "generation" (ie orig. to final.)

As you say, there may not be a perceivable difference. However, there is no "extra time" involved. You either render to a proxy and then swap (this is instantaneous) or you render an intermediate. Actually, a lightweight proxy should be faster to make than a good quality intermediate and be faster /more responsive to edit.

Your choice of course.

I mention this only because there is an awful lot of misunderstanding about this on the internet. It's a good skill to have.