Goodbye ND filters? Adding 180 degree shutter look in post

[Deadcode]

2 hours ago, kye said:

I think the point that @Deadcode was making was that the extra frames can help the motion estimation better, which makes sense, although I'm not sure that it would actually work that way in how they would have likely implemented the video processing pipeline.

It works, just tested

24p timeline, 120fps footage, corrected amount of blur, and it's a beauty

[fuzzynormal]

1 hour ago, Deadcode said:

It works, just tested

Using ReelSmart Motion Blur?  If that plugin is effective, consistent, and fast I'd consider it over using ND filters.  Costs about the same as an collection of cheap filters.  Man, I'd love to ditch my variable ND's.

Would like more real-world testimonials though.  Anyone else?

[Deadcode]

17 minutes ago, fuzzynormal said:

Using ReelSmart Motion Blur?

Ultrapro pure scientific test

GoPro Hero 7 @ 2,7K 24/30/60/120fps 

Resolve, 23,976 fps timeline, Reelsmart motion blur, blur corrected to framerate

of course frame drop can occur, no retime/frame blend correction

To my eyes the 120 is very nice, but the difference in this scene is less then what i expected.

[Video Hummus]

1 hour ago, Deadcode said:

Ultrapro pure scientific test

GoPro Hero 7 @ 2,7K 24/30/60/120fps 

Resolve, 23,976 fps timeline, Reelsmart motion blur, blur corrected to framerate

of course frame drop can occur, no retime/frame blend correction

To my eyes the 120 is very nice, but the difference in this scene is less then what i expected.

What was the processing time like and what CPU and GPU did you process it with? 

[Video Hummus]

1 hour ago, fuzzynormal said:

Using ReelSmart Motion Blur?  If that plugin is effective, consistent, and fast I'd consider it over using ND filters.  Costs about the same as an collection of cheap filters.  Man, I'd love to ditch my variable ND's.

Would like more real-world testimonials though.  Anyone else?

Same. I know Christian Maté Grab has said he ditched Various-NDs and used Reelsmart Motion Blur. The processing times seem bad though with his iMac Pro. Maybe Macs don’t get GPU acceleration ever since Apple dropped support for OpenCL.

Hopefully we will see Metal 2 support or even a built in solution from Apple in the next FCPX release.

I have heard Apple is finally giving their pro apps some love, and indeed, they recently just released a major update for Logic X.

[Deadcode]

9 minutes ago, Video Hummus said:

What was the processing time like and what CPU and GPU did you process it with? 

110 sec processing time for the 52s long 1080p 24p h264 clip with NVENC acceleration.

So around 13 fps processing speed.

Without RSMB the processing time is 14 s, so this effect is very taxing even in 1080p.

Config: i7 3770, 16GB, GTX1070, RSMB 5.0, Resolve Studio 16.1

[gt3rs]

Is RSMB much better that the motion blur in Resolve Studio? 

I may need to test it out the free trial.... price is fair if it is much better than resolve motion blur filter

The biggest advantage for me is to take frame grabs as photo and with higher ss is easier to stabilize or slow down with motion estimation...

[Deadcode]

16 minutes ago, gt3rs said:

Is RSMB much better that the motion blur in Resolve Studio? 

I may need to test it out the free trial.... price is fair if it is much better than resolve motion blur filter

The biggest advantage for me is to take frame grabs as photo and with higher ss is easier to stabilize or slow down with motion estimation...

It's almost the same comparison as Neat Video vs Resolve's own noise reduction.

It gets the job done waay faster than a 3rd party plugin, but the results are not the best. Worth the trade off? Depends on your needs. I ditched Neat years ago, but i prefer rsmb over the built in solution

[kye]

6 hours ago, Deadcode said:

It works, just tested

24p timeline, 120fps footage, corrected amount of blur, and it's a beauty

I know it would work.  My question is if the motion tracking uses the extra frames for motion estimation or if they get thrown away before the tracker is applied.

The reason I care about this is I recall the demo from Lok where he did star jumps, and at 25p there was a frame where his arm was sloped down, next frame they were about level, next frame was more than 45 degrees up, next one was vertical.  

In that example, would the motion estimation pick up that his arm is moving very quickly?  or would it just think that there are several unrelated objects appearing and disappearing?   If it was filmed in 120fps then there would be 5 times as many frames and the fact it's one object that's moving would be obvious to the motion tracking.

5 hours ago, Deadcode said:

Ultrapro pure scientific test

GoPro Hero 7 @ 2,7K 24/30/60/120fps 

Resolve, 23,976 fps timeline, Reelsmart motion blur, blur corrected to framerate

of course frame drop can occur, no retime/frame blend correction

To my eyes the 120 is very nice, but the difference in this scene is less then what i expected.

They all look quite similar to me, but the difference between 24 and 120p isn't that much as both of them appear to have approx a 45-90degree shutter.  If you compare the size of the blur with how far the object moves in the next frame then you'll see that the blur isn't that long in any of these modes.

I know I'm not that sensitive to SS in general, but my impressions of these kinds of tests are that if you have a bit of blur then it mostly looks fine.  It's when you're filming in direct sunlight and the SS is like 1/2000 that the motion effect breaks down and it looks more like a burst of unrelated images, or something moving but somehow moving through a medium that gives it strange sharp textures as it travels.

I guess the whole point of this is to do away with NDs, which would mean that shots would range from being indoor and 360 shutter to being in direct sun and being 1/4000.

[colepat]

I agree with @kye. I'd love to see some stress tests on outdoor filming with a crazy high shutter.

[fuzzynormal]

3 hours ago, kye said:

the whole point of this is to do away with NDs, which would mean that shots would range from being indoor and 360 shutter to being in direct sun and being 1/4000.

1/4000 would be typical in mid-day sun.  I'll shoot some video tomorrow and apply the trial plug-in.  See what happens.

Usually have a bunch of ravens flying around....filming that should provide a decent stress test.

[tweak]

That test looks promising.

[seku]

that reminds me of talk we had ages ago 😄

back then i was chastised for going for motion blur in Resolve. Filming at a higher shutter also an edge when stabilizing footage... as you skip the motion blur from that... you have to deal with rolling shutter tho.

Just apply motion blur after stabilizing

[Evgeniy]

I did my tests a while ago and decided that it wasn't worth it. 

RSMB provides pretty good results when you have a simple object moving against a simple background, but in challenging situations like shooting from a moving car, it will give lots of artifacts. Also, rendering times are horrendous. It's going to be much faster to use fixed NDs then to apply RSMB in post.

[gt3rs]

Did some better tests with horse showjumping (complex bg with foliage, fences and sharp turns of the horse) at 1/400 and adding motion blur in resolve is not that great too many artifacts I prefer the video without motion blur. I need to try the plugin to see if the result is better.

[projectwoofer]

Has anyone any recent info on how good/accurately works adding motion blur in Davinci Resolve nowadays? Vs that other plugin maybe? I don't intend to throw away my ND filters but a 90 or even 45 degree shutter angle would greatly help with stabilization or optical flow in post.

[kye]

oops..