Salto Motion Capturing Suit

Hello community! I found an extremely interesting project the other day on kickstarter, backed it and just wanted to spread the word in case you are interested. It is a full body motion capturing suit which looks perfectly affordable and usable for developers in need for specialized animations. I'm sure you are aware that currently Mixamo is giving out their really high quality character animations for free, and there are LOTS of them. But there is still animations missing, like for example animations for a slingshot.

With the salto suit I believe making these would be not only possible, but rather easy and even fun. I just wanted to add that I am NOT affiliated with the project in any way, just a backer that thought the Unity Community is a good place to tell about it. It is already fully funded, so coming on board now would probably be a good idea .

Anyways, here is the link: https://www.kickstarter.com/projects/2132274632/salto-captures-your-every-move-vr-ar-film-vfx-gami

Enjoy!

 

Thanks for sharing this does look neat. I'm a bit skeptical that I didn't see a camera anywhere and they are doing this solely with MPU's (even though they look like they are the newest MPU's from invensense the 9xxx series that have sensor fusion right on the chip). I've played with this chip in the TI sensortag and while it seems to be much better than what any of the current phones have, it is still prone to drift. Now with 19 of them and known places in the suit they can probably do some IK stuff to even calibrate between them, but it certainly would still need to get the animations tweaked after. At that point wouldn't a multiple camera motion capture setup be preferable, especially for indies? Maybe I'm missing something....

 

Thanks for your kind words about the system! As you highlighted already, we have been using the 9xxx-series from InvenSense. The sensors are pretty robust, but unfortunately not entirely perfect. This is one of the challenges we are dealing with right now, trying to obtain higher accuracy and minimize drift, especially the accumulated one that occurs when positioning. But we have a good solution for the Salto which we will be testing in the upcoming weeks before we finalize the hardware architecture. We do not have cameras, but it would be a good addition to the system if you are looking for absolute position. Many studios have contacted us and see potential in previs and even recording of movements for production as there is almost no cleaning compared to a optical system.

 

So motion capture is usually like $100,000+ and these guys came out with a suit for $700?

...

And it works with VR?

...

Well, I guess my sixense system will be used primarily for things that make sense for being operated with a handle in vr, like mechs, tanks and guns. Also, damn! They got funded and the price is going to be $700!? AHHHH

I can't react to this now, I'll be back in ~13 hours to try again. Thanks.

 

It is hard not to be sceptical especially after the issues PrioVR and Sixense have been having. I probably won't back the project but I do wish it luck.

 

We understand that people are super concerned due to previous experience with Kickstarter campaigns. Although we believe we differ from these companies in many aspects, which is easy for me to say when people are sitting back with lost funds. I have backed dozens of campaigns and at least 3 of them has failed, but it does not really stop me from backing others, as i truly believe we can't determine peoples failure on others failure.

And thanks!

 

I just checked the site for Perception Neuron to see if the were also experiencing delays. The site says that due to manufacturing delays, pre-orders will begin shipping in October. I'm not sure if the site just hasn't been updated and the orders are shipping or there is another delay.

Salto is half as expensive than PN seemingly, although it doesn't seem to come in a one size fits all exoskeleton kinda thing like PN, but instead actually clothing that needs to be properly sized. So it could end up being either more expensive or more time consuming if you want more than one person to be able to use it. Though I'm sure @mzamani can address that since he is in the thread.

As a VR controller these suits might work okay. But I think as far as replacing an animator for other types of game animations they might at best save a little money. I haven't really been able to find out how much cleanup of the animations is required for PN, but from my understanding even the big expensive and more accurate rigs require a bit of cleanup. Many of the free but unrefined MoCap sets out there illustrate how crucial that step is. I would still love to have such a suit for prototyping purposes, since the cleanup can come later.

 

We are actually making the textile in various sizes, as "One size fits all" does not seem ideal with what we have planned in terms of comfort and flexibility. We have close contact with animation studios which are already planning the system in their workflow, especially as a replacement for manual animation and pre-visualization. The clean up is really minimal, but of course the technology is not perfect due to the nature of IMU's, but we are doing what we can on a software level to minimize the accuracy error, and we feel that we reaching a point where we can say that the quality is acceptable.

 

Sixense has been delayed since july 2014.

How is this suit powered? Battery pack on the torso? Each node having it's own little batteries? I hate battery operated technology

 

The entire suit is powered by one battery source via USB on the what we call the Salto Hub, which basically is the gathering point for all the sensor strings. We do not have batteries in each individual sensor!

 

Usb you say? Can the suit be operated while plugged in for motions that don't require the whole body to turn? There are a lot of VR applications where I only need the use of my hands/arms.

 

The suit can be supplied with a single USB source and data connectivity either via WIFi or USB. Selective data stream is also possible!

 

I meant will it consume battery while plugged in? y'know, cuz I hate batteries

Also that's great that it can do selective data streaming. Any chance a future model will be sold piece by piece so you don't need the entire suit on? I don't mind the wires all over me still to connect to the central hub, but if I'm capturing hand gestures it seems silly to still wear the entire suit.

Great product either way, will probably buy after release so you guys make $700 instead of $550

 

@mzamani, that's pretty impressive! Looks like you are having a very solid approach. I am looking forward to it.

 

Tomnnn:

Thank you!

You can either choose to power it with a battery pack or which ever USB power you have! We have been discussing a more modular system, where you do not need to buy the entire system. More on this will come in the near future.

 

I am truly glad to hear, i hope you will follow us

 

Now I'm interested and setting some money aside. I have a $2500 bonus coming in november. If the modules can be hooked together to work like the full suit, I'm sold! (when said system is available, of course)

So, any big companies threaten you since you're undercutting them 99.4166666667%?

 

I realize for professional motion capture studios the cost is high, but are you aware of the other options for indies, even available right now? Basically, the cost is about equal to iPi Soft Basic edition and 6 motion cameras for a 20 x 20 capture area. To me the cost seems about right and certainly not drastically different than other indie type solutions (which only require a little animation cleaning as well, I don't think any solution will get away from requiring some cleaning).

I certainly wish the project the best of luck and hope you guys do great things. I'm a little concerned about MPU only but maybe you guys can crack the code that so many others have failed at. If you guys do I'm sure this will be a hit but I by no means look at this as a sure thing.

 

Not sure if this has been answered already, but what format does the system export to? The reason I ask is because I am using Maya LT and it doesn't support plugins or BVH files, as far as I can tell, so I'm wondering if this would be an issue. Any way to support Maya LT for indie devs?

 

The bonus is more than enough for a suit

Not really, they are looking at us, especially the optical systems, but we have not been threaten.

 

We have some great contacts within the indie game industry, the small ones as well as the big ones. They are all seeking a solution where they can sit at their desk and test movements as well as record mocap data for their animations. This is something that is almost impossible with a optical setup, for starters you have to sit in a pretty restricted environment and usually you only have one system for this purpose. With the suit you could have one on each workstation.

Thanks a lot! We are putting a lot of effort in choosing the correct hardware architecture and creating proper algorithms to increase the accuracy and stability of the sensors. We have used the system in real time before, so we are very confident about cracking the nut!

I hope you will follow us on our journey.

 

@cyberpunk: Unfortunately i am not sitting right next to the Maya guys right now, and its the first time i have heard about Maya LT. But do you know if it even supports any kind of mocap data, such as FBX since BVH is not the case? Furthermore i understand that Maya stand-alone has a BVH importer, but we need to look closer into this for Maya LT.

PS: The forum thinks i am spamming, so i can't quote you!

 

Cameras is bads for everything because occlusion And I looked for 'motion capture cameras' and still saw the cheapest was ~$2,000.

Nice one

 

That's why you have 6 camera's so you are never occluded from all the cameras. Asus xtions (depth sensor) is around $80 and if you use camera's instead Ps eye cameras are around $10. I've used Ipi software before and it does produce good results for motion capturing animations on a couple hundred dollar budget.

That being said if you want real time I understand.

 

The multi camera setup is by far not as flexible as such a suit. There are some interesting applications that can not or hardly be achieved with optical solutions.
- More flexibility for walk or run cycles (tricky when there are only a few cameras)
- Capture the motion in the actual environment
-- E.g. in a train to get the natural movement (not sure whether the sensors could handle that )
-- Or in a crowd (chances are high at Comic-Con that you blend in nicely with such a suit)
- Capture a huge amount of people at the same time

The prototyping potential could also be huge. I could easily just experiment with animations and apply them to the characters, adjust the state machines, check what's missing or what's not good enough, capture the animation again or add another one. This could allow outstanding iteration times.
For that, the animations just need to be good enough and could later on be replaced by better ones. The amount of flexibility alone sounds like a dream to me! There is the potential that I could do that alone, at home in my small office!

 

That's all very true, I was simply responding to the cost factor for an indie level setup which has long since not been anywhere near $100000.

As said before I hope this product does actually succeed, the main benefits as you and others mention would be for prototyping in real-time, and not needing to stay in a 20 x 20 area. They definitely have a coding challenge on their hands to compensate for the drift. I am mainly skeptical since John Carmack (obviously a much better programmer than any of us) has stated before (paraphrasing) that for GearVR/Oculus that they toyed with all the possibilities MPU's, cameras, etc. and that they were unable to achieve their goals with MPU's alone and they had access to special prototypes. New frontiers are crossed every day though and I'm sure we all hope that they can solve the problem.

 

I found perception neuron to be really bad for mocap and this looks awfully similar (unless I'm missing something).

Also it doesn't appear that you're tracking the head, based on that video.

 

Do you have Perception Neuron? I remember seeing it at Siggraph in Vancouver and thinking that it looked rather jittery but that was a while ago.

But that leads to a good question. How is this different from priovr and perception neuron?

 

We are tracking the head, you have seen some examples where the head band is not attached, but it is a part of our standard system. Could you please explain why you find PN bad for mocap?