Is there a point in using QuickTime Player's 'maximum' quality for streaming Hi8 analog tapes?

cjp987 wrote:

I just compared a video captured with "Maximum" quality to the same video compressed with the H.264. The compressed version looks way better. The original is grainy while the compressed looks very crisp. I guess there is a lot of correction that goes on during the compression that makes it look better. They both suffer from what I guess you would call ghosting or doubling of the image when there is movement. I thought this is what deinterlacing was suppose to correct.

QT Player "Maximum" (PAL, I guess NTSC is slightly different) is the original tape resolution 702x576 with very large ProRes codec so it is meant for further editing and re-encoding.

QT Player "High" (PAL) is somewhat "worse" 585x480 resolution already with small H.264 codec so it is meant for delivery.

As you discovered, QT Player deinterlaces both so there is that "ghosting" in moving scenes.

The old QT Player 7 "Device Native" is the original tape resolution 720x576 (yes, also 702x576 is original but I won't go to tech details here) with largish DV codec so it is meant for further editing and re-encoding. But it preserves interlacing so there are interlacing "stripes" or "ghosting" or "neither visual glitch" in moving scenes depending how it is viewed (in VLC etc) or deinterlaced and re-encoded for delivery (in Handbrake, Shutter Encoder, Final Cut Pro, ffmpeg etc). Naturally I prefer the last option (and do a "bob" deinterlace which preserves both interlacing fields in 25 fps PAL and outputs it as double frame rate 50 fps). The new QT Player makes that impossible because it deinterlaces as "ghosting" method when importing no matter what. So I personally would avoid that and use other apps to import DV/D8 tapes (old QT Player 7, old iMovie v4-6, Final Cut Pro etc).

cjp987 wrote:

So what happens if you take an H.264 encoded video and do further editing (just clipping or splicing of video clips). Does the decode/re-encode process further degrade the video?

H.264 is a delivery format not really meant for further editing but that has not stopped people doing that.

Yes, H.264 can be losslessly be cut and trimmed with straight cuts, but only "to the GOP" or group of pictures that usually happen at about 1-8 second intervals. Apps like QT Player cut to those GOP I-frames and just hide the non-cuttable I&P frames that some apps like Avidemux can then reveal.

Any further editing needs lossy re-encoding. And repeating that leads to "generation loss" and even more quality loss.

On the other hand, the much larger ProRes files can be edited frame accurately because it is, erm, an editing format with much less compression so it is not meant for delivery.

Interlacing (and rectangular pixels, and NTSC 30/1.001 ≈29.97 frame rate) are a nuisance from the past when ancient cathode ray tube TV's and technology still needed such clever tricks.

Interlaced footage has two temporally different lines with a field rate of 50 (PAL) or 60 (NTSC) that together make frames at half of the field rate. But in moving scenes they are different and if you put them on top of each other you get that ghosting or stripes depending how you view it.

New TV's and computer screens are not interlaced but progressive so those ugly interlacing lines must somehow be removed from the same frame. One common option is just to delete the other line and maybe interpolate its contents. The other "Bob Weaver Deinterlacing Filter" method is to preserve both lines, interpolate and double the frame rate with slightly smoother motion.

iMovie v1-6 has a setting to split scenes into smaller clips if there is a break in the timecode (i.e. the recording was halted). If that setting is turned off, iMovie v5-6 no longer has a maximum clip size of 2 GB (=9 minutes 27 seconds) previous versions had. I guess some versions did not add a file suffix .dv to its Media files (you can just grab the imported DV movies from the iMovie project folder or ctrl-click and open the project package later versions use, and archive them).

Various iMovie 1-6 versions had various irritating bugs but I had workflows to avoid them. Let me know if you plan to use some version so I can dig a memo for it.

In the past I used MPEG Streamclip (great app that runs up to Mojave) to deinterlace and export as compressed .mp4. So one great and easy option is to use that and call it a day because MPEG Streamclip handles the gory details like converting rectangular DV pixels to square pixels etc behind the scenes very nicely and accurately.

MPEG Streamclip can drill also into the newer iMovie project package, open the tiny reference .mov that contains the movie the user has edited, and export THAT as .mp4. Very handy.

But MPEG Streamclip does not support bob deinterlacing and I felt bad about throwing 50% off from the input so I now use ffmpeg for that and IMO the quality is good and better than with FCP (but not much better compared to MPEG Streamclip output with much less fuss).

Other options are to use Handbrake or Shutter Encoder (I have not tried to use VLC for this) to convert .dv to .mp4 but there are numerous options to choose from while MPEG Streamclip has a cleaner interface. Shutter Encoder can also run cookbook-style readily made ffmpeg commands via its GUI which is great for people who are afraid using the Terminal.

Importing as .dv or .mov with the DV codec is lossless (*) compared to the original DV/D8 tape so just use that for archiving (* bad tape spots and other hiccups while importing are flies in the ointment so there might be a few dropped frames per hour, though).

cjp987 wrote:

"start a new clip at each scene break" option, but that only seems to impact Digital8 tapes. Hi8 tapes start a new clip at 1hr no matter what the setting.

Yes, originally analog footage does not have those timebreaks in DV stream that can automatically split scenes into clips. The same happened when I used my old Sony TRV320E D8 to digitize my old VHS tapes and when I borrowed the camcorder so a friend so she could digitize her Hi8 tapes on-the-fly with that D8 camcorder to iMovie.

I have archived all my original VHS and D8 tapes as max 2 GB (max 9 min 27 sec) clips as an old habit and because larger clips are clumsy to handle (especially in the past with slow HDD drives). So pick anything between 10-65 min as a clip size. I'd not split even D8 footage into numerous small scene clips but edit and trim them later, if necessary (trimming as straight cuts and exporting as DV is lossless although in some iMovie versions it was not lossless with certain workflows I learned to avoid). Welcome to the rabbit hole of movie editing.

By straight cuts I mean simple trims and cuts with no pixel edits that would need re-encoding.

That artifact in the bottom of analog footage is normal analog "head switching" that was usually hidden in old TVs with their "overscan" feature.

Because old tape technology was not that exact, the same happens in both sides of at least VHS footage due to "blanking etc". The actual PAL image area is 702x576 although the whole rectangular pixel PAL frame is 720x576. There are several slightly different ways to scale those rectangular pixels to square pixels (so round objects remain round on a computer screen) but I have used the figures from the article below by scaling 4:3 PAL 720x576 to 788x576 and the cropped that to the final 768x576. With 4:3 NTSC scale 720x480 to 648x480 and then crop to the final 640x480. With 16:9 different numbers must be used.

DV/D8 footage does not have those analog artifacts and the image area covers to whole frame including the sides. So using those standard scaling and cropping will slightly crop real image. So the user could omit the final crop but that yields a nonstandard resolution that usually does not matter much though.

http://web.archive.org/web/20140218044518/http://lipas.uwasa.fi/~f76998/video/conversion/

cjp987 wrote:

With 4:3 NTSC scale 720x480 to 648x480 and then crop to the final 640x480.

I understand you are trying to crop the artifact at the bottom of the image

No, that relates to converting rectangular DV pixels to square pixels which computer monitors use. If that is not done, then circles will not be circles but ovals in the output.