This forum uses cookies
This forum makes use of cookies to store your login information if you are registered, and your last visit if you are not. Cookies are small text documents stored on your computer; the cookies set by this forum can only be used on this website and pose no security risk. Cookies on this forum also track the specific topics you have read and when you last read them. Please confirm whether you accept or reject these cookies being set.

A cookie will be stored in your browser regardless of choice to prevent you being asked this question again. You will be able to change your cookie settings at any time using the link in the footer.

[HELP] Alternative over BasicVSR++ for getting best video results

I am using BasicVSR++ from Vapoursynth -> Others-> BasicVSR++(NTIRE 2021(3)) + Vapoursynth -> Frame->Resize->(AnimeSR, RealEsrgan) which I am getting the best results but its very time taking process.
Could please someone help me out with an alternative over BasicVSR++ which gives the best results with more speed?
Depending on what you are aiming for, there is no alternative to BasicVSR++ to blindly suggesting anything is hard and what I suggest might be helpful or totally wrong. Big Grin
Depending on what you are doing, combining filters like:
  • (conventional)MC Temporal Denoise with Settings set to 'very high'
  • (conventional)SpotLess with temporal radius 3 (wouldn't go above 5)
  • (ai)DPIR/DPIRDenoise (mlrt)
  • (ai)SCUNet
  • (ai)DPIRDeblock/DPIDeblock(mlrt)
  • (conventional)SMDegrain with high (5 0000+) SAD values
  • (conventional)KNLMeansCL with higher values
  • (conventional)HQDN3D with adjusted values
might help, but it really comes down to your sources, other filters might be better suited or not.
You might have to use some masking and custom filter orders, but some stuff BasicVSR++ does other filters simply don't.
Yes, it may be possible to write custom scripts to do some of it, but that would require a good understanding of writing Vapoursynth scripts and video editing and you would have to adjust your script to each source.
So you might not be able to get a similar result.

Cu Selur
Do I need to try all of these at once. Could you please provide any screenshots or sample script code so I can understand properly?
Actually, I was trying to do the 2 hrs animated movie to upscale using BasicVSR++ and AnimeSR but almost it's taking  2-3 days to pass 1 video conversion so that's why I am trying to find an alternative to little bit speed up my task.
You can use Vapoursynth Preview to look at the effect of the filters.
When filtering animes, it often comes down to how many small details you count as defails and not as noise.
Adding some line sharpening, darkening and/or luma sharpening might also help.
With filtering, it often comes down to personal preferences to me blindly suggesting anything specific doesn't make sense.

Cu Selur

   for Anime upscaling I suggest to use "Resize->RealESRGAN->realsr-anime" and to enable "Tensor RT" and "FP16" and set Streams=2.
   This should increase significantly the speed.
   I don't understand why you are using also "BasicVSR++", most of video cleanup is already performed by " realsr-anime".
   If you need to increase the sharpness of movie, for anime (only for anime) you can use: "Sharpen->aWarpSharp2" with Threshold=128, Blur=2, Depth=20



   At the following link you can find an album to compare the filters:
   The "Base" image is just the frame resized with "spline36"
   It easy to see that for anime BasicVSR++ is not introducing any significant improvement, ESR realsr-anime is able to introduce a significant improvement.
   It is interesting to note that aWarpSharp2 with Threshold=128, Blur=2, Depth=20/64 is able to introducing a nice effect on the image (this effect is good for anime, for movies the effect is awful).

   I hope that this can help.
Hello Dan,

Thank you for your reply. Do I need to use it with the QTGMC or just these settings will work?
And let me give a try for your solution.
Also for those movies which not animated which settings will be preferred over BasicVSR++ cause i am always supposed to do with it and facing the long time conversion ratio to achieve the goal of mine.
Personally, I don't like the line tinning effect of aWarpSharp. Smile
The question always is: What the input really is and what does one prefer.
There is no: "this method will work fine for all cartoon/animes" and different sources, require different filtering Most super resolution filters assume a rather clean source. Wink

Quote: Do I need to use it with the QTGMC or just these settings will work?
QTGMC is for deinterlacing (and if used on progressive content, in Hybrid under Denoise, to clean-up interlacing residuals), it's not for cleaning or super resolution.
If your source is not progressive, you need to convert it to progressive content to use most super resolution methods.

Cu Selur

Forum Jump:

Users browsing this thread: 1 Guest(s)