Nailing down 720x480 capture settings

Topaz Video AI General
1

I intend to re-capture my source content before using Topaz. I’m looking at 720x480 footage, and as I capture it, I have the option of encoding via H.264, AV1, or HEVC. Will Topaz be impacted much by my selection? Is it more bitrate dependent?

I gleaned Topaz encoder supported pixel formats but not decoder formats.

Side note: I’m capturing Rec. 2100 (HLG) to replicate the analog color. It’s a bit overkill, but 709 & 601 produced awful results. This can be formatted as:

  • P010 (10-bit, 4:2:0, 2 planes) :star: leaning towards this unless Topaz has changed since this point
  • P216 (16-bit, 4:2:2, 2 planes)
  • P416 (16-bit, 4:4:4, 2 planes)

The following post was really interesting, but a bit outdated, and since it’s locked I couldn’t ask any clarifying questions sadly: https://community.topazlabs.com/t/topaz-video-ai-v3-0-9/38102/107?page=6

Main takeaways:

  • TVAI(3.x.x) seems to be internally processed in rgb48le(R:G:B 16bit)? @TicoRodriguez
  • TVAI only supports 2x or 4x
    :arrow_forward: any suggestions for 2/4x my import resolution?

Again, my biggest source of confusion stems from whether to encode captures at H.264, AV1, or HEVC prior to using Topaz. I’m not targeting lossless archival grade footage, but I would like to retain some decent color grading this time around.

1 Like
2

The following warning message may appear when TVAI is executed with the CLI.

No accelerated colorspace conversion found from yuv420p to rgb48le.

This means that there is no acceleration for yuv420p to rgb48le conversion.
This means that TVAI’s internal colorspace is rgb48le.
VEAI 2.x was R:G:B 8bit, but from TVAI 3.x it is R:G:B 16bit.
The same is true for TVAI 4.x.

For enlargement, only 2x or 4x is supported.
Other enlargements are performed by FFMPEG’s built-in scale filter.