Author Topic: How to convert interlaced video to progressive with double the frame rate?  (Read 1010 times)

mark.uk

  • Newbie
  • *
  • Posts: 1
Hello! :)

I've been digitising my VHS collection to mpeg 2 files. They are 720 x 576, 25 frames per second (PAL) and interlaced.

Some of these are TV shows and I don't want to lose that smooth interlaced motion when uploading to Youtube, so I'd like to convert them to a 50 frames per second progressive video. And they'll need to be upscaled to 720p60 too, since Youtube only has 50/60 frames per second for HD videos.

Could someone guide me on the best way to do this?

Thank you!
Mark

eumagga0x2a

  • Moderator
  • Hero Member
  • *****
  • Posts: 4063
Deinterlace with VDPAU if you are on Linux with NVIDIA or else with the Yadif video filter, choose "Bob, temporal and spacial check" to double the frame rate. Use x264 as encoder and MP4 as muxer. Save a short fragment of the video to check that everything is fine. If the result is okay regarding deinterlacing, add the swsResize filter to upscale it to 960x720 without preserving aspect ratio if they are 4:3 or to 1280x720 if they are widescreen. Don't resample to 60 fps, leave them at 50 fps.

You will have to re-encode audio too, choose AAC. Use either the FDK or the LAV encoder.

edit: If YouTube insists on 16:9 aspect ratio, upscale to 960x720 and add borders (160 px left, 160 px right, zero top and bottom).
« Last Edit: April 04, 2019, 09:04:08 PM by eumagga0x2a »

coolgit

  • Newbie
  • *
  • Posts: 7
Got the latest nightly build and used yadif as usual. Bob, temporal and spacial check is missing and now only temporal and spacial check. What happened to Bob?

eumagga0x2a

  • Moderator
  • Hero Member
  • *****
  • Posts: 4063
"Field" = "Bob"

coolgit

  • Newbie
  • *
  • Posts: 7
Thanks.

Was it necessary to change from Bob to field? What possible achievement is there except wasting users' time.

eumagga0x2a

  • Moderator
  • Hero Member
  • *****
  • Posts: 4063
The assembly code in Yadif was incompatible with Microsoft Visual Studio C++, resulting in the filter being much slower than with gcc platforms. This was fixed.