Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Higher resolution slow motion workaround for Raven

This is why I love REDuser! So many knowledgeable professionals all in one place. Thanks for the clarification. So then in reality the "project Timebase" that is set on the camera really is irrelevant. 24fps will play back at 24fps no matter what the project timebase is.
I'm surprised none of the REDucation teachers caught that, in the course thats one of the first things they taught us was how to set the Project timebase. Which I now see if pointless as it is just metadata. Damn :( I thought we were on to something! LOL
 
There are many types of optical flow software. From simple plug-ins to much larger systems.

They are all based around calculating vectors and movement, basically tracking a grid of pixels through the frames. Once you know how the grids of pixels are moving then you can create in-between frames.

The finer the grid the higher resolution of movement detection. This technique is also good for increasing or adding motion blur too.

But the system falls down when you have opposing movements, if you have an arm sweeping through the frame left to right whilst the camera is moving opposite then the local movement of the arm becomes difficult to track. That's when applications like NukeX et al come in handy, because you'll need to separate the elements and slow them separately. However panning through a chain fence with motion on the other side would be a nightmare for most systems. In these cases you'd shoot the elements separately and combine them. Slowing rotating objects is difficult too (as in spinning around coming towards you). Most flow detections are based on straight vectors, not curved.

So in summary optical flow is very scene dependent. If there is just camera motion across a fairly station scene then it works very well. As soon as there are opposing motions then you need very high resolution grids, but even so there will be tearing and other odd artefacts. The better the application the more algorithms to throw at it (Nuke has quite a few options and some great recent optimisations)

Test first, understand where it will fail and you're good to go.

cheers
Paul
 
Listen to your reducation teachers. Set your project timebase and leave it at that. Do not use TIMEBASE as a way to do SLOWMOTION.

Read.

http://www.reduser.net/forum/showthread.php?144268-Slow-Motion-with-RED-Best-practices

David


Cameron McKinlay said:
This is why I love REDuser! So many knowledgeable professionals all in one place. Thanks for the clarification. So then in reality the "project Timebase" that is set on the camera really is irrelevant. 24fps will play back at 24fps no matter what the project timebase is.
I'm surprised none of the REDucation teachers caught that, in the course thats one of the first things they taught us was how to set the Project timebase. Which I now see if pointless as it is just metadata. Damn :( I thought we were on to something! LOL
Click to expand...
 
Last edited:
point is basically.

120 frames is 120 frames no matter what TIMEBASE you shoot it at and if you do it at 60FPS you just make a lot more work for yourself.
Use more lights, use the right shutter speeds. Fundamentals will get you nice images.

David
 
David Battistella said:
Listen to your reducation teachers.
David
Click to expand...

Thanks for the input David. Unfortunately, the teachers at REDucation weren't the most knowledgeable individuals, One of the "teachers" barely knew how to use the camera. He was a VFX artist. Hence why I posted it here.
 
Cameron McKinlay said:
Thanks for the input David. Unfortunately, the teachers at REDucation weren't the most knowledgeable individuals, One of the "teachers" barely knew how to use the camera. He was a VFX artist. Hence why I posted it here.
Click to expand...

I get that this kind of thing can happen. I will just stick by this. 120 frames is 120 frames. If you do a whole production at a 23.98 timebase and then switch that timebase to 59.94 (60P) and record at 120 all you are doing is really making the post task VERY difficult. You can not squeeze more out of those 120 frames by changing timebase. It's better to let those 120 frames play back in a 23.98 timebase where you won't have mixed timebase issues which can be a bit of a nightmare. If going between 23.98 and 25FP is a large pain in the ass.

Don't mess with the project timebase and don't mix them.

David
 
David Battistella said:
I get that this kind of thing can happen. I will just stick by this. 120 frames is 120 frames. If you do a whole production at a 23.98 timebase and then switch that timebase to 59.94 (60P) and record at 120 all you are doing is really making the post task VERY difficult. You can not squeeze more out of those 120 frames by changing timebase. It's better to let those 120 frames play back in a 23.98 timebase where you won't have mixed timebase issues which can be a bit of a nightmare. If going between 23.98 and 25FP is a large pain in the ass.

Don't mess with the project timebase and don't mix them.

David
Click to expand...

May be the explanation was too simplified.... I don't know if you can throw one timebase into another timebase, but what I believe he meant was to throw the 120p into 60p timebase; export the footage... Then throw the exported footage into a 24p timebase and add optical flow in the 24p timebase. At which point, I have to ask why not do some optical flow in the 60p timebase too, before exporting?
 
There is no need to change timebases when shooting higher FPS. That is for cameras that do not have native slomotion built in like the RED camera. It's real easy.

There is ZERO advantage to captureing 120frames at 60P compared to 23.98. ZERO.

It's not a good idea and it's threads like this that create problems for people.

120 frames is 120 frames in 60P or 23.98. It's real simple.

David
 
Kabilen Mullaithilaga said:
I don't know if you can throw one timebase into another timebase, but what I believe he meant was to throw the 120p into 60p timebase; export the footage... Then throw the exported footage into a 24p timebase and add optical flow in the 24p timebase. At which point, I have to ask why not do some optical flow in the 60p timebase too, before exporting?
Click to expand...

Because overall, it is a pretty useless idea. mess with framerate but do not mess with TIMEBASE and understand that there is a LARGE difference between the two.

David
 
Yeah this makes no sense. What is the optical flow doing? What is it interpolating? the original framerate of 120fps determines the relationship of the frames to reality and the playback framerate determines how much it is slowed down, but any intermediate framerate should have no effect. Slowing it down 2x and then 2.5x is the same as slowing it down 5x...

As for frame interpolation, it is highly content dependent. It is not magic. If you interpolate frames for content where a detailed foreground subject moves in front of a detailed background you will get distortion artifacts because it is interpolating background pixels that were obscured in previous frames so it has no data to reconstruct from. This is minimized for scenes where the motion in the frame is relatively slow compared to the framerate so that the difference between frames is small. If you want your super slow interpolated footage to look good make sure it's shot against a blurred or low detail background so the blurry halo around the subject in motion is less visible.
 
Noah Yuan-Vogel said:
Yeah this makes no sense. What is the optical flow doing? What is it interpolating? the original framerate of 120fps determines the relationship of the frames to reality and the playback framerate determines how much it is slowed down, but any intermediate framerate should have no effect. Slowing it down 2x and then 2.5x is the same as slowing it down 5x...

As for frame interpolation, it is highly content dependent. It is not magic. If you interpolate frames for content where a detailed foreground subject moves in front of a detailed background you will get distortion artifacts because it is interpolating background pixels that were obscured in previous frames so it has no data to reconstruct from. This is minimized for scenes where the motion in the frame is relatively slow compared to the framerate so that the difference between frames is small. If you want your super slow interpolated footage to look good make sure it's shot against a blurred or low detail background so the blurry halo around the subject in motion is less visible.
Click to expand...

Again, I think you guys are confused.
120fps is 120fps slowed down to 24p it is 5X the speed
That is not the part that you guys are confused about...

In any NLE, you dont' really need optical flow, twixtor or any other slowmotion plug-in to slow it down from 120p to 24p. 20% of the speed is 24p.
What people want when using optical flow, twixtor or the like is 1000 frames per second and up. If you're happy with 120p, 240p or even 300p (which is from what I see in the manual, a RED camera can do), then leave it alone... don't even play around with optical flow, twixtor or the like...

For whatever creative purposes that you need 1000, 2000 fps... seen in this video:
https://www.youtube.com/watch?v=FLMoJ7jl4sw&feature=related

Watch the entire thing, because the first few seconds of it is literally 60p or 120p...
You will see what 2000 fps when you get to the part where the bike flies off the ramp.

I think what is happening when you do throw it down from 120p to 60p, export it as 60p footage... you're just conforming it. Its a slowed down footage, then when you apply optical flow to make it 1000fps or 2000fps.... its just more smoother??? Again, I don't have proof, I will need to play around with it when I get the Raven.

So to reiterate, we are talking about frame interpolation, and it isn't magic...
BUT what we are trying to achieve isn't slowing down to 2X then 2.5X, instead 2X then 25X... which is the same as 50X... and in terms of frame interpolation 25X is better than 50X in one shot.
 
Last edited:
The 60P makes not difference. You are always only starting with 120 frames. whether it was at 60P or 24P make ZERO difference.
 
David Battistella said:
The 60P makes not difference. You are always only starting with 120 frames. whether it was at 60P or 24P make ZERO difference.
Click to expand...

Sorry, I edited my original post (the above) to add this bit:

So to reiterate, we are talking about frame interpolation, and it isn't magic...
BUT what we are trying to achieve isn't slowing down to 2X then 2.5X (which is the same as 5X), instead 2X then 25X... which is the same as 50X... and in terms of frame interpolation 25X is better than 50X in one shot.

 
Last edited:
Cameron McKinlay said:
its putting things in the menu system that change absolutely nothing that creates problems for people.
Click to expand...

Not taking sides here because I'm not able to separate what is being said to be sure of either statements. But I'm wondering if what David is saying is that when you go from a 24 fps TIMEBASE to a 48 fps TIMEBASE, you lower your available framerate for the format you are shooting... presumably you lower the framerate proportionally.

Yes, 48 fps TIMEBASE gives you twice as many images as a 24 fps TIMEBASE and people like Peter Jackson shoot that way to give a more realistic look to their movies when shown in a 48 fps exhibition.

But for simply getting a slowed-down set of images based on camera output only, a 24 fps TIMEBASE shot at 48 fps will give you the same amount of individual images as the 48 fps TIMEBASE shot at 48 fps. The only difference is the 48 fps TIMEBASE will not show up as slow motion.
 
Last edited:
You must log in or register to reply here.
Back
Top