Upsampling/Upconverting beneficial before editing footage?

matt_gh2

Would love to hear opinions on whether there is any significant benefit of upconverting/upsampling footage before editing and color grading. I'm relatively new to editing, so all opinions and knowledge is much appreciated. I'm wondering if GH2 1080P footage would benefit from upconverting it to 4k before editing...i.e. would it hold up better in grading and could it be pushed further before the image starts breaking down? Also wondering if it's possible(as talked about in Windmotion thread) and advisable to upsample GH2 1080P footage to 4:4:4 before editing...i.e. does this too help with better results in grading?

I edit in Adobe CS6 with access to all programs (Creative Cloud) such as Premiere Pro, After Effects, Speedgrade, etc.

BurnetRhoades

I'd say not for editing. Especially not 4K. There's no good reason, that I can think of, to make the footage bigger and fatter before you whittle it down. Pro cinema editors were delivering final cut EDLs after seeing the whole movie for six months or more at like 320 lines with the original Avid systems. Starting out with 1080/720 footage, like we have now, is like working with release print quality from the early days of digital non-linear. What's important in editorial is realtime.

All the extra detail doesn't really aid in the critical judgement calls of editing. That comes later. Notes about takes or sections that would be NG and not to be used for the edit should come from the logging and/or dailies/rushes or equivalent review of the footage prior to sitting down to really edit.

Your color is already being upsampled during this and the color grading process. Once your edit is locked you can replace the footage (and not all the stuff you may not be using) with specially processed versions of just the clips needed (Windmotion, 5DtoRGB, Film Convert Standalone, Neat Video, etc.).

Realtime is important for editing, not so much for color timing, so the big fat files are just fine here. Quality is now what's important. Realtime, at this stage, is a luxury that usually comes at a loss of quality so be careful.

LPowell

I often need to intercut footage from several different cameras taken at various exposure levels and color temperatures. My preference is to first do a technical grade on each individual take, normalizing all footage with consistent exposure and color balance. I find doing this minimizes mismatch problems when fading between cuts and makes it much easier to polish off an artistic color grade on the final cut.

I use Color Finesse in After Effects CS5.5 to do the technical grades. With Color Finesse you can save a complete grade as a preset which you can then quickly apply to similar takes. I export each clip as an AVI file using CineForm in High HD quality. This produces 10-bit 4:2:2 All-Intra video files that are about twice as large as Flow Motion v2 100Mbps files. The CineForm codec is available free with a GoPro CineForm Studio download.

An additional benefit from pregrading in After Effects is that it reduces the CPU load on Premiere Pro, and that makes it more responsive when scrubbing through the timeline. When you pile on a lot of filters and color adjustments to individual clips, Premiere will eventually start to bog down.

BurnetRhoades

@LPowell in CS6-PremierePro, MTS definitely outperforms ProRes on the timeline. I didn't use CS5.5, going to CS6 from CS4, where MTS on the timeline sucked.

But I don't apply any filters or color adjustments in Premiere, ever. I edit in Premiere and once the edit is locked import the PPro project (or FCP project if I'm getting an edit from my brother) and all filtration and color grading happens in After Effects.

The only benefit to doing any color manipulation, that I see, inside Premiere is the presence of scopes as a matter of fact and not something left up to plug-in developers.

LPowell

@BurnetRhoades I feel the same as you do about color correction in Premiere Pro. Unlike After Effects, Premiere has no color management, making it a poor environment for technical work. But for artistic grading, I really like Premiere CS6's new Adjustment Layers. In Premiere CS5 you had to flatten your edit into a stand-alone sequence in order to apply effects globally, and that was really an awkward workaround. Adjustment Layers in CS6 make it feel more like working in Photoshop.

BurnetRhoades

@LPowell yeah, I don't come from a heavy Photoshop background so it's only recently that I've come to appreciate them. The first time I did a project file with multiple adjustment layers applying global tweaks to a whole sequence, or section of a sequence, I thought back in horror to all those years of copy-paste.

maxmizer

I use 5DtoRgb, uncompressed, 709, full range, correct?

BurnetRhoades

I've been using the broadcast range for the transcode, since it matches what the GH2 shoots. I'll "un-ass" the footage, to use a favorite term a buddy of mine coined, inside After Effects to restore true black. There may be no harm in the other way but I recall being encouraged to keep this step using the GH2's native range by a fellow on the eoshd boards (aka:yellow) that sounded rather knowledgeable.

LPowell

@BurnetRhoades There's really no advantage to transcoding GH2 MTS files into studio-swing broadcast range. The metadata encoded by the GH2 into the MTS files flags them as full-swing Rec. 709, and when imported into After Effects, the levels extend across the full range.

Psyco

@LPowell

I thought GH2 MTS-files are 709 broadcast range and the best (direct) way with 5DtoRGB is using 709 broadcast range? This will alter the native data the least. Please correcte me if I'm wrong.

DrDave

I have occasionally used InstantHD for projects, it's not bad. Good for polishing a turd. Cineform is the best codec I have tried, I use it all the time. I don't use Lee's workflow but it looks better than what I use, which is upconvert to Cinform then edit in PP using AVCHD workspace.

On my copy of Premiere, the Cineform workspace does not work as good as the one from Adobe. Go figure. Maybe they have fixed that, I use the Cineform that I purchased, I haven;t tried the free version but if Lee is using it it probably is as good or better.

I have even on some cams gotten better detail than built in 24p by shooting interlaced and removing the pulldown with Cineform, which absolutely should not happen. Whatever. Slight difference in the aliasing for some reason.

@Psycho there seems to be some differences of opinion about superwites, 709 etc. I hasten to add before going down the rabbit hole that Cineform is highly configurable and you can change all that stuff in metadata. No rabbits wanted or required.

LPowell

@Psyco If you have After Effects CS5 or CS6, you can try it and see for yourself:

Open a new AE project and set 32-bit mode with the "HDTV (Rec. 709)" working space.
Import a GH2 MTS file, open its Interpret Footage dialog, and click the Color Management tab.
Check out the text: "Color values will not be converted because working space and footage color space match each other."
Now try a new project with the working space set to "HDTV (Rec. 709) 16-235"
Import the same MTS file and its interpretation will now read: "Color values will be converted from the assigned profile HDTV (Rec. 709)... to the working space HDTV (Rec. 709) 16-235..."

You can also use the luma waveform scope in Color Finesse to check out the exposure range in the imported MTS file. Since Color Finesse is not color managed, it displays the luma data as-is. With a dark frame in your first HDTV (Rec. 709) project, the blacks will extend all the way down to 0 IRE. This is also the case when you import the file into a project with color management turned off.

Here's the bottom line: If you use After Effects, it will import your GH2 MTS footage as full-swing Rec. 709 video, unless you deliberately configure it to remap your footage into broadcast studio-swing Rec. 709 16-235 color space. This is also the case with CineForm AVI transcoding, and I expect the same is true of 5DtoRGB as well.

BurnetRhoades

@DrDave "I have even on some cams gotten better detail than built in 24p by shooting interlaced and removing the pulldown with Cineform, which absolutely should not happen. Whatever. Slight difference in the aliasing for some reason."

Actually, this isn't as unusual as you might think, depending on the camera, codec and a lot of other factors involved but, essentially, you're trading spatial fidelity for increased temporal fidelity. With sufficiently clever technique you can regain some spacial quality from the extra temporal information.

This was much of the basis for the earliest form of Magic Bullet which created 24P imagery from interlaced DV that, in the course of conversion, was able to undo much of the damage done by both the compression and color sampling of DV25 recording, and why they suggested DV filmmakers go ahead an shoot in 60i/50i for conversion to 24P rather than any (usually bogus) 30P mode their camera might provide.

Add optical flow to the mix and you could very well get really nice 24P this way, potentially better than what the camera might give you with its own, native 24P shooting. It still might feel a little different from normal 24P in motion.

Psyco

Thanks @LPowell .

I remember a discussion about the GH2 changing its colour range when pressing record to the smaller broadcast range (which is visible in the histogram) - but thats just the display, isn't it?

So, I will use full range in 5DtoRGB from now on ;-)

Howdy, Stranger!

Categories

Tags in Topic

Top Posters