Does Hybrid have any tools/filters that can deal with the sort of flickering seen in this file?
https://www.swisstransfer.com/d/a7f3cc03...e21ce21064
I would like to be able to reduce/eliminate it and then possibly resize the file to 768x576. I appreciate that getting it to look decent at a higher resolution is probably a tall order given it is not the best quality source, though. Eliminating the flicker is my main priority.
The above is a clip but I've also uploaded the full file here since it is only small
https://www.swisstransfer.com/d/f104bd6a...95103556f5
Any tips/advice would be most welcome.
Thanks
J
Argh, overcompressed low resolution.
What graphic card are you using?
So I know whether machine learning based filters are an option
, those filters can sometimes really help a lot with such content.
NVIDIA GTX 1660.
I was having a play around with various filters and I found that Spotless 10 radius did make it flicker less (it's on the silver shelves in the background where you see most see the flicker). But I'm not sure if this is the right filter to use. It seems to cause occasional issues with other parts of the video e.g. hands appearing to disappear if moved quickly.
This is the clip with spotless rad 10 and resized to 768x576 using NNEDI3. Spotless isn't a good solution though due to the aforementioned issue with hand movements in some of the scenes.
https://www.swisstransfer.com/d/6455050d...7b0b0c63d7
The video also doesn't look as good/sharp as I'd like but I guess short of using AI tools, there's not a lot that can be done to improve the appearance of 480x360 compressed video. Any denoising/degraining just seems to make it look blurrier and I don't know what the best sharpen methods in hybrid are.
SpotLess with usually is a bad idea if you need a radius > 3.
The flickering is likely caused by bad deinterlacing + high compression causing aliasing.
Try something like: SMDegrain with really high values and Santiag
Code:
# removing grain using SMDegrain
clip = smdegrain.SMDegrain(input=clip, tr=6, thSAD=3200, thSADC=256, interlaced=False, opencl=True, device=-1)
# contrast sharpening using CAS
clip = core.cas.CAS(clip=clip)
# no resizing since resolution is already archived
# applying anti aliasing using santiag
clip = antiAliasing.santiag(c=clip, strh=3, strv=3, nns=2, qual=2, pscrn=2, exp=1, opencl=True)
also try the above and set SMDeGrain -> DCT to "use block DCT ...", that should help a bit more. (but will be a lot slower)
Cu Selur