AviSynth Deinterlacing Help...

[Shizam]

Alright, so I'm just a guy looking to put some 10 minute videos on Youtube. I got into AviSynth and became tempted by all of the available tools, so I decided to try nnedi3 and eedi3 combined for my deinterlacing. Problem is, it's way too slow for me. Looks amazing, but my compressor told me it would take at least 16 hours at 6000 bitrate and I simply can't have that. The most I'd be willing to deal with would probably be 10 hours... I have had good results with Yadif I guess, but I'm looking for something a little better with not TOO much added time. Any suggestions?

Oh, and I also can't have anything that will alter the frame rate. As long as there's an option to maintain my framerate, it's fine.

ALSO, when I say 16 hours, I mean with a 2-pass encoding. I've heard you can apply filters on just one pass, but is that wise? How would I do it, take out the filters in the AviSynth script for the first pass and put them back in the 2nd?

[jagabo]

There are some "fast" settings for TempGaussMC_beta1mod() posted somewhere around here. That is about five times faster than the default speed.

https://forum.videohelp.com/threads/319816-Improving-old-VHS-What-to-expect?p=1981816&v...=1#post1981816

Still about 1/10 the speed of Yadif() though.

[poisondeathray]

ALSO, when I say 16 hours, I mean with a 2-pass encoding. I've heard you can apply filters on just one pass, but is that wise? How would I do it, take out the filters in the AviSynth script for the first pass and put them back in the 2nd?

It's not wise to apply filters to only 1 pass (I would assume you mean the 2nd pass) , because filters can change the frametype distribution (e.g. It might have been an b-frame with the filter, but now it's a p-frame). For deinterlacing, it's even more drastic - it's a completely different source when you deinterlace so probably not even possible.

You can apply filters to a lossless intermediate first , e.g. huffyuv, lagarith, ut video codec , then use that as input to your compressor

You have to endure slow filters twice right now (they are applied on both passes), instead you can use them only once . On slow filters (like temporal filters) or stacked filters , even though you are technically doing 3 passes now, it ends up to be significantly faster when you use slow filters, since that is usually the bottleneck (not the actual encoding.)

[manono]

Or just run 1-pass CQ encodes. If it's going to YouTube the size doesn't matter as long as it's below their 2 GB (?) limit.