Welcome to our community

Be a part of something great, join today!

  • 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 :)

STEADXP on RED.

Björn Benckert

Well-known member
Joined
Dec 5, 2007
Messages
7,423
Reaction score
74
Points
48
Age
52
Website
www.syndicate.se
https://vimeo.com/255270817



Just got one of these little. Bought it used from a friend. Possibly the coolest 100 Bucks I ever spent on camera gear.

Not sure how it comes if you buy one new but I needed to get a sound cable, microSD card, smallrig horseshoe mount and a Gopro battery as the previous owner just sold me the little box with none of the above. So got the rest from amazon.

Please don't analyze this little test too deeply. A lot of things is not right here. I did a sloppy calibration, used a soft relatively long lens (helios 58mm oval iris pretty much wide open) and its running with a guessed redraw time of 15ms. So there is a lot that could be improved. Using a wider slightly more stopped down lens would make huge difference.

The zoom in factor from monstro 8kFF was 20%. So something like going from 8k to 6k... Some argue 6k is enough so easy enough to set your guides accordingly shoot and let steadxp use the look around to smooth out your shakes.

I have not gone too deep into the software, just clicking my way trough to get a clip out the other end. But thats family easy. I can imagine that there is a lot of things that could be done to improve things. Simply have different softness for pan tilt and roll would be cool. Normally I find it nicer ot zero roll, not give it any smothing at all but keep camera leveled around roll at all times then smooth tilt a lot and pan a bit less. Any how I assume that can be all done if digging deeper.

Now as RED got gyros in the cameras but is so far not implemented in red cine or the STK such. imagine if red would buy this code and simply let us smooth out all footage from epic and forward.

I for one will experiment with this further and I´m trying to get the guys at steadXP to let me access the gyro data so I can make nuke and flame plugins, but so far I got nothing but their encrypted stuff to play with which means I can only use it in their software. Not bad but for sure would be more preferred if it could work as a OXF script or such.

I think cool if more buy this little gadget and we can keep calibration settings for different camera/lens configs here someplace and share our findings.

Source before stabilization is applied.

 
It seems nice Björn, but wouldn't you get the same result with Warp Stabilizer ?

Kristin

No, the SteadXP records motion telemetry to help in the post process I believe

https://steadxp.com/

I have never had great results in software only post stabilization tbh
 
It seems nice Björn, but wouldn't you get the same result with Warp Stabilizer ?

Kristin

You mean like in premiere warp stabilizer or such? Nope, as this is not based on optical flow but on true camera rotation from a hardware gyro. So this thing can stabilize your shoot no matter what goes on infront of the camera. No pixel tracking or analyze needed. When using optical flow needs references points to be successful. If your warp stabilizer can not find a few points that gives info about movement then it will fail. This will do the job perfect even when you pan over a flat wall or drag everything out of focus and it knows how to keep the camera level as It calibrate on a flat surface.

Its basically doing what your gimbal does but after you shoot the stuff.

So basically yes, in terms of zooming into the shot and mapping it onto a sphere and pan around it does the same thing as warp stabilizer. But here the pan tilt roll data is hardware driven. It also comes with far more advanced settings for lens calibration and sensor calibration. So if set up correctly it will take your lens distortion into account and also your sensor rolling shutter issues for pan tillt roll. As it knows the rotation and you give it the redraw time of your sensor. I estimated it for monster to be (15ms/4320) for FF. I think thats close but things would improve if I get the exact number.


looks like this.

IMG_5635 by Björn Benckert, on Flickr

It got an internal battery and a little dual ended headphone cable that feeds a sync feed into two of the sound channels. That way it finds sync and use it to line up the roll, pan,tilt data with the shot.

Would be cool to record red with 100fps and sound as then we could use frame oversampling to stabilize shots using this thing. Like shooting 360deg 100frames stabilize and then merge two frames into 1 and drop two frames. That would give a very good 25p 180 deg with butter smooth stabilization and all the rolling shutter issues cut in half. Not that those are that visible but still.
 
Thanks for your answers, Björn and Zeb. I understand a little bit better now. The issue with this little box is that, if I'm not mistaken, you have to plug it to a sound input... Is there a way to make it work if you don't have any sound inputs with your Red camera ?
 
Thanks for your answers, Björn and Zeb. I understand a little bit better now. The issue with this little box is that, if I'm not mistaken, you have to plug it to a sound input... Is there a way to make it work if you don't have any sound inputs with your Red camera ?

I think you can also plug tc into it. And in theory I have not tried you should be able to manually sync it but have not tested.
Most red cams have sound input and I think this also works so you can have sound on one channel and this on the other.
 
What about the built-in gyro sensor of the RED? Will that be something that some day can be used for the same purpose (without the extra outboard gear)?
 
The new DSMC2 models don't have sound inputs, you have to buy a module for that and the sound quality of the preamps of it isn't good in my opinion. It's a limitation of the system if you want to keep this light and small. I'll look into it anyway, thanks.
 
No, the SteadXP records motion telemetry to help in the post process I believe

https://steadxp.com/

I have never had great results in software only post stabilization tbh

Automatic post stabilization is usually failing if the scene gets a bit complex with a lot of paralax and intersecting objects in the scene and such.
But if you use syntheyes or moca or such and learn how to do it properly. But its no easy task. It takes years to learn and is a profession where you can be better paid than even the best animators, so to learn you need to dive deeper than just watching the tutorials and hit analyze. So image based stabilization takes effort to get right.

Even with this thing you can improve the result quite far if everyting is done proper, lens calibration at a given focus, as focus zooms an a perfect redraw pixel line number is a must. As long those are even just a bit estimated or inacurate things deteriate.

Then you can also keyframe the shots to make the most out of the canvas shot. To me this should be all done in a compositing tool where you could set up a hirarcy of axises and also smooth x y and z seperatly.

The thing is we all have gyros in our camerasand they actually write the data into the 3rds together with the rest of the meta data then if the gyros are good and have enough resolution. Redcine could simply have a stabilize function that took your canon lens and everything into account. That would as I see it put red in a class by its own. None of the competitors have this especially not in their older models and they dont keep track of ef lens data on a frame basis.
 
Hello Björn,

If you do an affordable OFX plugin for Resolve using the gyro of the Weapon body metadata through their SDK I'll be the first one to buy it!
;-)

Pat
 
Hello Björn,

If you do an affordable OFX plugin for Resolve using the gyro of the Weapon body metadata through their SDK I'll be the first one to buy it!
;-)

Pat

Ditto. That doubles your sales right there!
 
I have one of these too, but haven't getting into testing yet. The few tests I did failed miserably and it's a bit of a hassle to get a good result. It's not like it's put on and works out of the box.
My initial idea was to use it for minor stabilization, like when using a hand camera but not moving too much. However, it seems that it's much harder to get it going and doesn't seem to sync well through the minitele solution. Feel like it's more of a toy for GoPro and DSLR users than a professional gyro tracker. I think SteadXP as a company has the wrong focus. GoPros get better and better internal stabilization and their GoPro edition doesn't work with the latest GoPro. I think they should ditch the skiing and skateboard approach for this gear and focus on professional application. This would go out of stock directly if it was aimed a little better at the filmmaking industry.
 
On shots with shallow depth of field focused on an actor, I find it almost impossible to get good camera solves for vfx background replacement(since all the trackers on the background are way out of focus, and the only thing in focus is the moving actor which I have to mask out from the camera tracker). This seems about perfect for problems I am having with shallow depth of field camera solves, thanks for the testing.
 
I have one of these too, but haven't getting into testing yet. The few tests I did failed miserably and it's a bit of a hassle to get a good result. It's not like it's put on and works out of the box.
My initial idea was to use it for minor stabilization, like when using a hand camera but not moving too much. However, it seems that it's much harder to get it going and doesn't seem to sync well through the minitele solution. Feel like it's more of a toy for GoPro and DSLR users than a professional gyro tracker. I think SteadXP as a company has the wrong focus. GoPros get better and better internal stabilization and their GoPro edition doesn't work with the latest GoPro. I think they should ditch the skiing and skateboard approach for this gear and focus on professional application. This would go out of stock directly if it was aimed a little better at the filmmaking industry.

Yes this is far from plug and play... It takes atleast a few minutes to set up :)

Compare it to a gimbal. Where you need special cables, and preferably set it all up the night before as you don't know how it will balance with all the gear, focus motors, mattbox, batteries, exoskeletons to carry it all, etc. Then when you swap the lens all needs to be recalibrated.... Then jump from the gimbal to a crane and back... then most ask for a second camera head.

With this thing you can put it on camera. Plug in the sound cable and if you got a green blinking light you are good to go. Al else can be done after the shoot. The AC and dit can shoot the STEADXP grid for calibration and the rest is taken care of in post.

And when using a normal gimbal stabilizer it´s always a bit of a balancing act to not set it too smooth or to stiff. So this will do wonders even when using a normal stabilizer. As it provides an option to smooth the moves even further. Or if we get the data readout, remove smoothness or alter it where needed. Knowing pan tilt roll is huge.

But sure it could have an aluminium chassi. It could have a few buttons like record and calibrate. Instead as of now its made out of plastic and as soon as you stick the cable in it, it will start to record.

Not sure if its possible already but external power from USB or Dtap would also be a good thing. It has usb now but I think that is more for charging the internal battery and offloading files. To me a short usb cable straight to red usb power out would be great.


But off course. We want the internal gyros data that we already have in the r3d´s to come to use. If red can not write the steadxp part into redcine, steadxp should be able to read r3d´s and make it possible to stabilize r3ds in their program.

Some one need to organize a "professional steadxp" group buy. Where it comes in titanium chassi of preferred color. :)
 
Also one more thing to consider. Your normal gimbal does not help you with rolling shutter. If you pan it fast or tilt, then things will skew. With this little guy and the right calibration it will not.
 
On shots with shallow depth of field focused on an actor, I find it almost impossible to get good camera solves for vfx background replacement(since all the trackers on the background are way out of focus, and the only thing in focus is the moving actor which I have to mask out from the camera tracker). This seems about perfect for problems I am having with shallow depth of field camera solves, thanks for the testing.

Yes or try stabilize the camera using optical flow if you film into a dancing crowd...
 
Does it change the geometry of the picture (Squew)?

I very much use the rolling shutter time to correct what is going on. The stead Xp cal file ask for your cameras resolution and also the time it takes per row of pixels to be redrawn. So in this test, as I did not know better I got 15ms as somewhat right for red camera sensors to redraw and then I decided it by the hight in pixels for the FF monstro sensor.

Now, I can tell that my shit is not dead on. As they even get better results with go pro cameras. But thats easier as the lens is then very much known and the redraw time is also a known. I asked Graeme for the redraw time but he does colors not exposure times so he did not know and told me to ask suport. I figure I do that first thing monday morning and see if I get an exact number. I think if it´s 100% correct then it will very much improve the jelly effect.

Also, I know from stabilizing using image based stuff. it´s super crucial to get a perfectly flat image. The better lenses you use the easier it is. But lenses that breaths when focusing is simply altering their amount of fishbowliness and FOV when altering focus. So best is to calibrate using about the same focus as you use when shooting. For most lenses that means you need to shoot a huge check chart. As steadxp got a screen driven check board best must be to use a large screen, like a large flat panel TV or such and shoot the calibration take on that with like 3 meters distance and sharp focus. Then also your lens calibration would be more accurate.


But the coll thing with this is that you don't really have to have anything calibrated when you shoot, its all added in post. So as long as you remember what lenses you where using you can shoot the cal maps afterwards and also refine those in post. And to me even my really shitty calibration take that took seconds and was done using my laptop it still worked quite well.

Then they got some issues with the calibration being linked to the resolution of the file. I guess thats due to the rolling shutter things. But to me it should be possible to make so that it does not matter if you stabilize the 2k proxy, the 4k proxy or the full 8k stuff. As its all the same math behind them just the resolution differ. I mention it to the steadxp guys today but got no reply on that matter so far.

r3d´s straight in would be something. Then they could just charge us for the software and use the internal r3d gyro.

Also for all gimbal makers out there. I think, for next NAB. Maybe time to start record your gyros onto a memory card and send sync to camera. A gyro stabilized image with possibility to alter the stabilization in post. Would make your gyro better than all the other gyros.
 
I have given up the hope of the built in gyro on dsmc1 and maby even dsmc2. But i have hi hopes for hydrogen will have seriously good gyro onboard for this.

Your test looks very promesing Björn. How fast is it to stabilize in post? Warp stabilizer on a 8k clip is not so snappy. Also the few times i like the warp stabilizer is when converting a shoulderrig shot to static. Funny thing is you always have to turn of the warping in the warp stabilizer cause it always looks ugly.
 
Back
Top