HDRx motion interpolation

[Ruairi Robinson]

If this "more natural motion blur" thing is a computer interp/best guess, will there be situations where it breaks down and guesses wrong? From my experience, every attempt at this, ever, has broken down in difficult situations.

like say.. fields of grass blowing in wind. translucent objects. Etc. Stuff that has rapid subframe motion... light streaks on rapid/jagged camera movement.

Actually it doesn't take much to crash into the limits of computer interpolated vector mblur. So I am pretty skeptical.

Has the hard stuff been tested? Does the short snapshot frame use the longer exposure as a guide for vector motion or some other new fangled magicnology?

R.

 

[Jannard]

If this "more natural motion blur" thing is a computer interp/best guess, will there be situations where it breaks down and guesses wrong? From my experience, every attempt at this, ever, has broken down in difficult situations.

like say.. fields of grass blowing in wind. translucent objects. Etc. Stuff that has rapid subframe motion... light streaks on rapid/jagged camera movement.

Actually it doesn't take much to crash into the limits of computer interpolated vector mblur. So I am pretty skeptical.

Has the hard stuff been tested? Does the short snapshot frame use the longer exposure as a guide for vector motion or some other new fangled magicnology?

R.

If you shoot HDRx properly (normal exposure and under exposure) you have lots of options with two tracks of conjoined images. My bet is that you can find a way to break anything if you look hard enough. So what are your options?

If the standard HDRx blend doesn't give you the "look" you are after (Magic Motion), even with the slider control to use as much or little of the under exposure you want/need... then you have The Foundry's MNMB (More Normal Motion Blur) to work with.

If you have shot something that breaks that, and a small percentage of situations might, then turn off HDRx and live with 13 stops you get natively from EPIC.

Lots of options here. Remember that, just as with shooting RAW or a film negative, processes improve over time. My expectation is that motion estimation will also, particularly since there is now a good reason to focus more on it.

Jim

 

[Peter Matthes]

Lots of options here.
Jim

 

[Philipe Ratton]

nice !

 

[Ruairi Robinson]

If you shoot HDRx™ properly (normal exposure and under exposure) you have lots of options with two tracks of conjoined images. My bet is that you can find a way to break anything if you look hard enough. So what are your options?

If the standard HDRx™ blend doesn't give you the "look" you are after (Magic Motion), even with the slider control to use as much or little of the under exposure you want/need... then you have The Foundry's MNMB (More Normal Motion Blur) to work with.

If you have shot something that breaks that, and a small percentage of situations might, then turn off HDRx™ and live with 13 stops you get natively from EPIC.

Lots of options here. Remember that, just as with shooting RAW or a film negative, processes improve over time. My expectation is that motion estimation will also, particularly since there is now a good reason to focus more on it.

Jim

Sure. Just the words "motion estimation" give me cause for concern. A lot of little things that happen subframe, are things that really bring an image to life. Can't fake that.

R.

 

[Jannard]

Sure. Just the words "motion estimation" give me cause for concern. A lot of little things that happen subframe, are things that really bring an image to life. Can't fake that.

R.

Ruairi,,, did you actually read my post? Just curious. Have you seen the Magic Motion clip? If you are worried, why not shoot without HDRx? Or try it and ignore it if you can't get the results you want? I'm not sure what your point is... I'm all ears.

Jim

 

[Ruairi Robinson]

Ruairi,,, did you actually read my post? Just curious. Have you seen the Magic Motion clip? If you are worried, why not shoot without HDRx™? Or try it and ignore it if you can't get the results you want? I'm not sure what your point is... I'm all ears.

Jim

My point is that you are screaming 18 stops 18 stops, and the fact that it may come at a cost to motion quality is going to get lost in the hype. Because "most people won't notice the difference". And maybe they won't but this is something I look very carefully at. And it's not a tradeoff I find thrilling. Great dynamic range or accurate motion? Pick any one! you make this choice sound much easier than it may end up being. Who wants to give up one over the other?

Or maybe it will look awesome, but the clips posted thus far hardly put it through its paces. I look forward to seeing footage that proves me wrong.

I want to be wrong.

Yours cautiously optimistic,

R.

 

[Cüneyt Kaya]

hi ruairi, i like your work and the humorous posts you make, very entertaining.
but this one simply wasnt nice (the way how...)
what about waiting and just see for yourself if it works when its in your hands and if its fits your needs.
if not, you still get the best digital cam on the planet.
not a bad situation at all for filmakers.
be positive

 

[michael zaletel]

How can adding an extra image data track have ANY downsides other than file size or reduced max frame rates? If you are shooting 24p 5K and aren't hurting for disk space, why not get more in the can in case you need it? Are there any scenarios anyone can think of where you would actually be better off not shooting HDRx?

Man, I want one of these to start shooting with.

-Michael Zaletel

 

[Jannard]

Ruairi... with all due respect, you have gotten used to an illusion. 24fps at 180 degree shutter is not how we see. Have someone swing their arm over their head from one side to the other. If this was shot 24fps at 1/48th shutter all you would see is a big motion blur until the arm stopped. But that isn't what your eyes see. They see a combination of motion blur and sharper references to the arm all along the way. We all got used to complete motion blur from a film camera because the compromise was made for average speed of movement at a normal distance. Magic Motion is much closer to what the eyes see. Don't just take my word for it. Many have seen it on the 40' screen here at RED. HDRx has other compromises in different areas. As you suggested, Magic Motion has a hard time keeping up with very fast moving shiney objects in a dark environment. MNMB works much better for those circumstances. Can you break HDRx? Sure. Just like you can pan too quickly with any 24fps camera. We have to deal with judder. We got used to grain to the point where we thought we actually didn't want to give it up. There are a ton of not so great things we get from every camera. Does that make HDRx a not so valuable tool? I don't think so. It is a terrific tool with many uses that are better than any we have ever had before at capturing high dynamic range motion. But we have to learn how to use it along with the tools being developed for it. If you want to focus on what it doesn't do, you just might miss what it does do better than any tool you have ever seen.

Jim

 

[Jannard]

Ruairi... one last thought. If you shoot HDRx you can use the "under" track for motion tracking. Then either combine that track with the "A" track and use a slider to determine just how much of it you want to use. Or... you can then ignore it altogether after you have used it.

Lots of creative options. Sorry if I sounded defensive... long day/week/month/year with not enough sleep.

Best.

Jim

 

[CJ Roy]

Jim-

Can we get more footage of HDRx in motion? I've only seen the Vegas strip shot of the firetruck. Seeing its limits would certainly help judge where it'll be best used.

Also, will the read/reset speed of the Epic MX sensor be fast enough that rolling shutter isn't a problem during the HDRx?

thanks.

 

[Álex Montoya]

It's a great tool and I can't wait to put my hands on it.

 

[Jannard]

Jim-

Can we get more footage of HDRx™ in motion? I've only seen the Vegas strip shot of the firetruck. Seeing its limits would certainly help judge where it'll be best used.

Also, will the read/reset speed of the Epic MX sensor be fast enough that rolling shutter isn't a problem during the HDRx™?

thanks.

Rolling shutter discussion seemed to go away with the RED ONE M-X. EPIC is twice as fast read-reset.

We are buried trying to build cameras and support these features. Will get some more footage out as soon as we can. Michael Cioni just posted something from John Schwartzman test... probably only got permission for stills. We'll work on it when we can. Need sleeeeeep.

Jim

 

[Ruairi Robinson]

Ruairi... one last thought. If you shoot HDRx™ you can use the "under" track for motion tracking. Then either combine that track with the "A" track and use a slider to determine just how much of it you want to use. Or... you can then ignore it altogether after you have used it.

Lots of creative options. Sorry if I sounded defensive... long day/week/month/year with not enough sleep.

Best.

Jim

No worries, thanks for the info.

I hope you understand when I ask questions like this, I am not attacking you. I want it work comprehensively and be awesome.

Best,
R.

 

[CJ Roy]

Rolling shutter discussion seemed to go away with the RED ONE M-X. EPIC is twice as fast read-reset.

We are buried trying to build cameras and support these features. Will get some more footage out as soon as we can. Michael Cioni just posted something from John Schwartzman test... probably only got permission for stills. We'll work on it when we can. Need sleeeeeep.

Jim

I know that the MX read/reset was initially supposed to be closer to 5ms versus the R1's 8-9ms. But that caused some problems, and the read/reset had to be pushed back to the 7-9ms range, correct?

I still get the rolling shutter with high frequency vibrations from various vehicle mounts, lightning strikes, camera flashes, sirens, some handheld, etc.

Anyways, glad to hear this is a non-issue for the Epic. Looking forward to getting mine... and hopefully some Epic-M's.

Thanks.

 

[Gunleik Groven]

R1 MX Read/Reset looks very similar to M currently.

There was talk about a plan re-set it to the shorter reset time, but it hasn't come out yet.

 

[Casey Green]

Ruairi... with all due respect, you have gotten used to an illusion. 24fps at 180 degree shutter is not how we see. Have someone swing their arm over their head from one side to the other. If this was shot 24fps at 1/48th shutter all you would see is a big motion blur until the arm stopped. But that isn't what your eyes see. They see a combination of motion blur and sharper references to the arm all along the way. We all got used to complete motion blur from a film camera because the compromise was made for average speed of movement at a normal distance. Magic Motion is much closer to what the eyes see. Don't just take my word for it. Many have seen it on the 40' screen here at RED. HDRx™ has other compromises in different areas. As you suggested, Magic Motion has a hard time keeping up with very fast moving shiney objects in a dark environment. MNMB works much better for those circumstances. Can you break HDRx™? Sure. Just like you can pan too quickly with any 24fps camera. We have to deal with judder. We got used to grain to the point where we thought we actually didn't want to give it up. There are a ton of not so great things we get from every camera. Does that make HDRx™ a not so valuable tool? I don't think so. It is a terrific tool with many uses that are better than any we have ever had before at capturing high dynamic range motion. But we have to learn how to use it along with the tools being developed for it. If you want to focus on what it doesn't do, you just might miss what it does do better than any tool you have ever seen.

Jim

Post of the year (+2010).

 

[Floris Liesker]

The idea of relying on a computer to guess subframe motion for me is a repulsive waste of time. Not a fucking chance.

Usually Ruairi picks his words very carefully when describing his feelings.

 

[Joe D'Arcy]

Everyone has a bad day sometime.