Proper toolchain for interlaced content?
 

Would appreciate feedback to see if I am doing something wrong with my interlaced tool-chain.

I'm working my way through transcoding NTSC content for hosting on my Plex serve and thought I would give one last honest attempt at preserving interlaced fields. Normally I just use QTMGC as I find combing artifacts very distracting.

The workflow I tired is an Avisynth script to do some field aware/safe filtering from the vob I ripped from DVD. Then I used the x264 encoder in VirtualDub to encode the video stream using the "--tff" option. Finally, I remuxed the video with some AAC encoded audio from an de-interlaced comparison video. MediaInfo confirms the following:
Scan type : MBAFF
Scan type, store method : Interleaved fields
Scan order : Top Field First

The issue is that this result is disappointing when I compare this to the progressive, de-interlaced alternative.
I have tried MPC-BE and VLC (disabled de-interlace) to review. The combing artifacts are still very noticeable and it appears neither player advance the video content by field instead of by frame. I would expect players to replicate interlaced playback by advancing by field at double frame rate instead of updateing the frames in a progressive manner. Worse, the the bobed video enjoys a 14% smaller stream size (probably in part from QTMGC's de-noising) and looks sharper!

It seems to me that with the neglect that interlaced content probably sees in development today, I'm better off using the best de-interlacer I can while transcoding rather than leave it to whatever fast algorithm is used at playback.

Assuming H.264 here, interlace streams poorly, and most apps/software don't properly understand it. So I'm not sure how far you'll get trying to set up a Plex server. MPEG-2 handles interlace properly, but then large files becomes the issue.

- "Smart" TVs are pretty dumb.
- And various streaming sticks -- Chromecast, Roku, Fire Stick, etc -- are not much better.

The best that can be done for such setups is proper QTGMC. Ideally with 59.94fps, but some content forces 29.97fps due to on-screen clocks/etc that confuse QTGMC bob.

When you look at content with VLC, run the Yadif-2X filter.

Overall some of your reasonings and understanding may be a bit off, but you finds are accurate.

Retaining interlace is mostly for archived master files, or broadcasting. Not streaming, even over LAN.

Remember: there's no rule stating you can only have 1 file. Keep an interlaced lossless (or high bitrate) interlaced max-res 720x480/576 sample. The create "watch copies" for viewing as H.264 deinterlaced, properly cropped (704x480), with overscan masked. That is most ideal.

How would the 704x480 be displayed as 4:3? Via an embedded flag in the file or would another conversion have to be done to get 640x480? Or would it be better to render it at 640 horizontal in the first place?
I suppose 704 horizontal is correct for DVD, but unsure about all the other devices.

Standard-definition video uses a magical thing called non-square pixels: http://www.miraizon.com/support/info_aspectratio.html

The video should contain an aspect ratio flag in its metadata, telling an obedient media player whether it's 4:3 or 16:9.

Does Hybrid allow for insertion of a 4:3 flag in its featureset?

Also, does 704x480 NTSC scale correctly the same way 720x480 does when given a 4:3 flag?

Hybrid uses PAR signaling: https://forum.selur.net/thread-597.html