Any news on possible upcoming compability of the TensorRT version with SVP and MPC-HC?
Got the Rife-4.6 Ensemble True/False Versions to work by simply downloading them and place them in the models folder, but my 3080 struggles with real-time 1080p interpolation (720p works fine), so I am glad for every bit of additional performance optimization.
Are there any clear instructions/clever workarounds to get the new seemingly more performant version to work with SVP or is it outright not possible at the moment (and why)?
Thanks in advance!

These limits are also discussed in the thread I posted above.
https://www.svp-team.com/forum/viewtopi … 288#p25288

MAG79 wrote:

"Uniform mode = switch between uniform interpolation (1mmmm) and Scene changing.
Adaptive mode = switch between uniform, 1m, 1m50% and Scene changing.
When we set m1 and m2* limits value to 1 we get intervals: 0..1 for uniform, 1..1 for 1m, 1..4000 for 1m50% and 4000..max for Scene changing. Thus we have 1m50% algo enabled for all good scenes in movie. SAD<1 it is not achievable for real material.

*m2 limit is only called m2. It fact it is limit for 1m50% algo. It is undocumented feature."

and answered why it is 1 and not zero there:
https://www.svp-team.com/forum/viewtopi … 285#p25285

"what s the theory behind the change to 1?"
"Because it close to 0, but 0 leads to SVPMgr crash. I found it today and I will fix it in SVP 3.1.3."

So maybe "adaptive with m1 and m2 limits set to zero" is correct and was just done that way in development history, but there is some kind of additional script/code which defines the timing of the MI frames? Because the results I am getting are matching the tooltip description, and not just adaptive with limits to zero.

The code in the mentioned avs-file from gaunt looks like this:

m33 = MSmoothFps(super, backward_vec, forward_vec, finest=finest, num=FramerateNumerator(src)*5, den=FramerateDenominator(src), algo=algo, ml=ml, sadml=sadml, sadgamma=sadgamma, block=block, blend=blend).SelectEvery(5,2)
m67 = MSmoothFps(super, backward_vec, forward_vec, finest=finest, num=FramerateNumerator(src)*5, den=FramerateDenominator(src), algo=algo, ml=ml, sadml=sadml, sadgamma=sadgamma, block=block, blend=blend).SelectEvery(5,5)
m50 = MSmoothFps(super, backward_vec, forward_vec, finest=finest, num=FramerateNumerator(src)*5, den=FramerateDenominator(src), algo=algo, ml=ml, sadml=sadml, sadgamma=sadgamma, block=block, blend=blend).SelectEvery(5,3)

(wid_den==1)&&(hei_den==1) ? last : eval("
 m33 = m33.BicubicResize(src.width, src.height, b=0, c=0.75)
 m67 = m67.BicubicResize(src.width, src.height, b=0, c=0.75)
 m50 = m50.BicubicResize(src.width, src.height, b=0, c=0.75)
")

interleave(src,m33,m50,m67)
smooth_video = SelectEvery(8, 0,1,3, 4,6)

Further, MAG79 stated:
https://www.svp-team.com/forum/viewtopi … 569#p26569

"Any news yet on the 1.5M/1M50% mode?
It was some technical issues and I had to stop my experiments. In few days I think I will return to this topic.

As for now we have 4 variants of 1.5m:
- 1m11m (it is gaunt's idea, not ready, stuttering is visible)
- 1mm1m with 50% distance (ready to use in SVP 3.1.2, more smoothness)
- 1mm1m with custom distances (ready to use in SVP 3.0 with gaunt mod, more smoothness)
- 1mm1m with 50% intensity (it is gaunt's idea, I'm work on it, even more smoothness)"

Maybe it helps to clear this up and to find a way to change the distances of MI-frames in SVP 4.

Oh, maybe I misunderstood something here, but why says the tooltip in the SVP menu otherwise?

"1,5m": similar to the 1m mode, but the interpolated frames are closer in time to the original frames, which reduces the visibility of artifacts."

My question is, if it is possible to change that "closer in time" to different values, in order to balance smoothness and artifacts, somewhere between 1,5m and 1m.

The tooltip description above also seems to match with the results if I check the interpolated video frame-by-frame (little difference between MI frame and last original frame, bigger difference to the following original frame).
"Uniform on good frames and 2m on all others" sounds like a extrem form of adaptive mode, but when its implemented that way, shouldn't I notice the difference in smoothness between good frames and difficult ones? 1,5m seems to output MI frames very constantly over all scenes. If I switch to 2m, it's much less smooth than 1,5m even in difficult scenes, but if I switch to uniform, it's a lot smoother than 1,5m even in good scenes (?).

I am using a DLP projector and AVR, there is no possibility for me to pass on 72hz to the projector, unfortunately I am limited to 1080p@60hz.

Hello,

SVP Developer MAG79 explained in this old thread (2012) https://www.svp-team.com/forum/viewtopic.php?id=962 the 1.5m Mode.

"Original 1m algo looks like: 1 m33 m67 1 m50
1m50% looks like: 1 m17 m83 1 m75
If you look to gaunt's avs file you will see: 1 m37 m73 1 m55
m33 means 33% distance from left original frame to right original frame.
You can customize these distances in gaunt 1m algo."

How can I customize these distance values for the 1.5m mode in SVP Pro 4?

I would really like to experiment with that, because the 1.5m mode gives me no visible/irritating artefacts, but is of course less smooth.
I want to try if I can tradeoff a little more distance of the interpolated from the original frames (and so so little more artifacts) for a better smoothness, but still with less artifacts than 1m or uniform mode.

Thank you for your answers.