1K ws at 400fps suprised and pleased

[Jaakko Rinne]

I have to disagree about Twixtor and 50% speed. When Twixtor manages to calculate a good frame between two original frames (for 50% speed), it'll just as likely manage to do 5, 10 or 20 good frames. And when it can't do that one good frame it won't do any good ones. The difference in image content between the original frames is what matters, not what speed you set Twixtor to.

 

[Brice Ansel]

I have to disagree about Twixtor and 50% speed. When Twixtor manages to calculate a good frame between two original frames (for 50% speed), it'll just as likely manage to do 5, 10 or 20 good frames. And when it can't do that one good frame it won't do any good ones. The difference in image content between the original frames is what matters, not what speed you set Twixtor to.

I'm not a Twixtor expert, I did achieve good results until 50% speed, if trying with the same sample to go under the 50% having good results has always been near impossible for me.
Best
B.

 

[Jaakko Rinne]

It depends on the content: fast, erratic movement might give better results at 50% as you might not notice the fact that the new frame is missing movement in some axis. But my personal experience has been with pretty much all optical flow style retiming tools that when you get one good new frame you can get a dozen and when you can't get one you can't get none.

EDIT: Just to clarify: I'm talking about getting a good estimation between two original frames, not for whole clips. When slowing down to more than 50% you really want to pick the footage you use. With 50% speed only every other frame is new so an errant frame might not be noticeable. With 10% speed the one errant frame will be 10 errant frames so any problems will obviously be pronounced.

 

[Brice Ansel]

That's a good tip Jaako, thanks for the clarification.
So when you are working on a whole clip how do you isolate the suppose bad frame then?
And how do you get rid of it?
Do you put a key frame on it with 100 speed and then go back to Lower speed?
Or do you choose a clip without bad frame?
Best
B.

 

[Ian McGrew]

thank you for doing a "test" without shooting water haha

 

[Jaakko Rinne]

Or do you choose a clip without bad frame?

The easiest way is to choose a clip (or the parts of a clip) that doesn't give you bad frames. And having footage that's shot with optical flow in mind always helps. Here's an article from 2006 that's still relevant ("The Rules of Optical Flow" section being the most useful part).

Essentially if you want to do a major speed change in post (10% speed or maybe even 1% speed) you want to help out the motion compensation by separating the background and foreground with masks. And you might do the effect on a long clip, but you'd only use the parts of the clip that look good and work your edit around that. If it's a clip where you keep going from 100% to 10% to 100% to 10% you'd time the changes so you're running at 10% in parts where you don't get bad frames and at 100% where you get them. And if you've got a really difficult shot, it's always best to calculate the cost of VFX hours vs. renting a camera that's capable of higher framerates.

 

[Elsie N]

I'm not a Twixtor expert, I did achieve good results until 50% speed, if trying with the same sample to go under the 50% having good results has always been near impossible for me.
Best
B.

So, what could Twixtor do with 1k 400 fps footage?

 

[Brice Ansel]

thank you for doing a "test" without shooting water haha

Hé hé....

The easiest way is to choose a clip (or the parts of a clip) that doesn't give you bad frames. And having footage that's shot with optical flow in mind always helps. Here's an article from 2006 that's still relevant ("The Rules of Optical Flow" section being the most useful part).

Thanks great article!

So, what could Twixtor do with 1k 400 fps footage?

Well, I guess it depend of the sample you want to work on.
I gave a try on the sample in my first post.
The specific last cut when the pigeon fly away, and I coudn't achieve good results at 50% speed.
The blury part of wings did fall into pieces with twixtor.
I'll give another try later with some masks to see if I can achieve at least 800Fps.
In the mean time if their is someone who's really familiar with Twixtor and want to play with the sample I did post on post 1,
you are more than welcome to download the file on Vimeo and post here your results.
Best
B.

 

[Jaakko Rinne]

You should note that when intending to do retiming to look like 800fps you should shoot at 1/1600 shutter.

 

[Rob Ruffo]

I have to disagree about Twixtor and 50% speed. When Twixtor manages to calculate a good frame between two original frames (for 50% speed), it'll just as likely manage to do 5, 10 or 20 good frames. And when it can't do that one good frame it won't do any good ones. The difference in image content between the original frames is what matters, not what speed you set Twixtor to.

Well no, because at some point you have a weird ping-pong effect, hair for example, in real life does not move between positions in a series of linear motions, as it will when twixtored. At 50% each other frame is "real" so weirdness is not as noticeable.

 

[Rob Ruffo]

The easiest way is to choose a clip (or the parts of a clip) that doesn't give you bad frames. And having footage that's shot with optical flow in mind always helps. Here's an article from 2006 that's still relevant ("The Rules of Optical Flow" section being the most useful part).

Essentially if you want to do a major speed change in post (10% speed or maybe even 1% speed) you want to help out the motion compensation by separating the background and foreground with masks. And you might do the effect on a long clip, but you'd only use the parts of the clip that look good and work your edit around that. If it's a clip where you keep going from 100% to 10% to 100% to 10% you'd time the changes so you're running at 10% in parts where you don't get bad frames and at 100% where you get them. And if you've got a really difficult shot, it's always best to calculate the cost of VFX hours vs. renting a camera that's capable of higher framerates.

Asier and cheaper would be to rent an Epic no?

 

[Jason DeSimone]

very nice

 

[Jaakko Rinne]

Asier and cheaper would be to rent an Epic no?

Indeed. Like I said, it's always best to calculate the cost of VFX hours vs. renting a camera that's capable of higher framerates. And this calculation should also include the value of better results from real overcranking vs. inferior results from retiming in post.

 

[Jaakko Rinne]

Well no, because at some point you have a weird ping-pong effect, hair for example, in real life does not move between positions in a series of linear motions, as it will when twixtored. At 50% each other frame is "real" so weirdness is not as noticeable.

If you'd read the whole thread before replying, you might have noticed that I'd pointed that out. Erratic or fluid movements will not work well with optical flow.

 

[Brice Ansel]

Just gave another try at 400fps with a much higher shutter then push to whole clip to 30% with Twixtor (aprox 1333 fps if my math are right).
Jaakko thanks for the tip.
Best
B.

 

[Doug Beatty]

fps = [1/.3]*400
1333.33

 

[Brice Ansel]

fps = [1/.3]*400
1333.33

Your math sound more acurate than mine.
Post fixed Hé hé.
Cheers
B.

 

[Mark Pedersen]

Ahhh Paris. I will be there in a month.

Thought I'd change the subject. Getting a little too technical here.....

And hey, no one commented on the Leica. I have had very similar experiences with my Vario APO 70-180. Leica's are amazing. Just got a 90mm Summicron ASPH and my god what a wonderful lens... And on EPIC... optical potential realized.

M

 

[Brice Ansel]

Ahhh Paris. I will be there in a month.
And hey, no one commented on the Leica.

Maybe some fellows don't want to see the second hand market of those babies going to high in price.
B.