Basic Deinterlacing for Video Enhance AI

Could a basic deinterlacing be brought into Video Enhance AI? As it would be wonderfully useful for NTSC live action footage I am trying to use the product with.

Video Enhance AI is not a video encoder per se, you should properly encode your videos first (deinterlacing, ivtc, deblend…) and then feed them to Video Enhance AI. At least that’s how I see it, the role of Video Enhance AI is different from that of a proper encoding chain.

The one thing that would be great and would work for you (and for me) is the ability for Video Enhance AI to directly process AviSynth script files. That way it would allow for the chaining of any kind of filters we want, upscaling and encoding at the same time. (With a bit more control over ffmpeg inside Video Enhance AI maybe).

Deinterlacing is actually a pretty complex process since there are a lot of factors involved. As such I’m not sure if it’s worth having it in VEAI (outside of some kind of external plugin/filtering process).

  • Field Order
  • Inverse Telecine (Film Sources)
  • How to handle mixed Telecine/Video (usually safest to treat as video but the film portions will have extra frames padded in that way).

For interlaced sources with VEAI, I usually make a progressive conversion in TMPGEnc 7 or Handbrake (TMPGenc’s inverse telecine is far superior to Handbrake’s especially if the source is animated) - I encode the result as high very quality HEVC encodes (for space saving) (usually CRF 17 or even less than 17 depending somewhat on source material).

I’m looking for a simplified solution that has decent results.

If Video Enhance AI can scale well, I just want it to also pre-process with a common deinterlaced function to see if I can get a one-stop solution. I rather not have to take time to run all the footage through various vendor’s solutions.

I paid for the product partially because I was hoping to streamline the process of getting NTSC SD footage to HD. Perhaps I bought the wrong product.

Thanks for all the suggestions though.

As thompsonl said, you can’t simply run a deinterlacer on your NTSC footage, there are a lot of factors to take into account. The simple difference between true interlaced and telecine sources means that you need an advanced understanding of those things and cannot streamline it to a simple “deinterlace” checkbox or something.

I would suggest looking at AviSynth and its various plugins, since everything is free. QTGMC for pure interlaced material, and TIVTC + TDecimate for telecine content. Hybrid video should usually be encoded VFR.

I stand by what I said though. The ability to load an AviSynth script in VEAI would be a big leap and could achieve what you want. You would still need knowledge of AviSynth and its filters, but you would get a better handling of those things than software like Handbrake provide, without having to encode your video before sending it to VEAI.

I’d like to simply run a deinterlace on my footage. I don’t care if it’s perfect - give me a choice of a few deinterlacing methods – I need a clean front end for a tool like AviSynth or the old Donald Graff filters.

If I have to do all the work of running AviSynth manually then I don’t want want to throw money at an easy solution. Was hoping to pay to get simplicity.

I agree though to your idea: “The ability to load an AviSynth script in VEAI would be a big leap”

Any Topaz engineers willing to introduce such a hook / preprocess feature? Would help I would think.

Well that’s the thing, it’s not about giving a perfect result but giving even an acceptable result. Try running a deinterlace filter on a telecine source and see if you like that result. And assuming a wrong field order is probably even worse-looking.

If you’re looking for a front end for AviSynth, maybe you should take a look at something like StaxRip? In the end you will still need to be knowledgeable about AviSynth. Personnally I wrote my own video encoding toolchain to avoid manually writing AviSynth scripts or extracting VTS etc.

On the subject of loading AviSynth script well, VirtualDub does that and is open source, perhaps the Topaz team could look into it.