Motion Capture Testing - day #1

We've always known that, for the concept that we're working on to really work, we really needed to have motion captured animations.
There's no getting away from it... hand animations can only take your project so far and to really get that wow factor (I hate that expression) you can't beat MoCap.

When we started all this nonsense (some 5 years ago) MoCap was a pipe dream... it was something we'd "find a way to do somehow" ...either someone would have to give us a bunch of money or we'd "borrow" some time on an established MoCap rig, somehow.

However... that unfortunate situation is no longer true.

In the indie world, Words like "inaccessible" and "not for us" are finding themselves in less frequent use when referring to tools and solutions.
This website for example (and the fact that I'm able to blog on it and talk at you poor unfortunate souls) is testament to that very fact.

So, these last two weeks we've started looking at an indie friendly MoCap solution.
The one we've opted to evaluate is the offering from iPi Soft (pronounced "Eye Pie Soft" I believe ...Mmmm, Pie).

There's a lot of unknowns with this... and whilst it's accessible to indies (i.e. its not $10,000 dollars a day to use) it's not cheap either (more on that in a moment) so we're talking small steps here to see if it's right for us (and my bank balance frankly).
I thought it might be useful and even inspirational to some of you if I documented our experiences with this as much as possible ...hence this blog, there's more to come in future, but here's the story so far.

It's a markerless system... you don't have to wear a skin tight suit covered in ping pong balls (although it isn't specifically prohibited for those of you that might be so inclined). Instead, the product uses footage captured from between 3 and 4 identical cameras and, after a bit of calibration, tracks your motion by seeing *you* and not any of the aforementioned ping pong balls.
That's my somewhat oversimplified technical understanding of what's happening ...fortunately it's not my job on this project to have in-depth technical understandings of much, I leave that to much smarter people than me.

So, today (Saturday 28th of August) we had our first test run, it wasn't entirely successful (yet), here's why...

Myself and one of the aforementioned smarter people got together in his well lit office space and rigged up 4 playstation 3 Eye cameras on tripods and atop wardrobes etc. (PS3 Cameras are the camera of choice for this product it seems, but you can technically use any 3-4 identical webcams).
The PS3 cameras don't have any kind of "proper" fixing system to keep them in place where you put them and they have inflexible leads coming out the back ...I'd recommend getting yourself some blu-tack or duct-tape if you decide to follow in our footsteps, it worked well for us in that regard.

With the cameras in place we plumbed the whole lot into Andy's (smarter person) main gaming / development rig and fired up the capture software.
We had a feed from the cameras immediately (always good) and then with a minimal amount of tweaking we had a really good feed at a steady 30 frames a second.


The next step is the calibration process... here's where it didn't yet work.

You need a small bright torch to move around the zone that you're capturing in and the calibration process tracks that torch movement and identifies its limits.


Turns out choosing the best lit venue we could for this was the worst thing we could've done. No amount of window closing, hanging towels across the windows and closing the blinds etc. could keep the (atypically) bright English sunshine out of the room and the cameras couldn't adequately track the torch amongst all the other visual clutter it was picking up.


So, from here then...

Andy is going to try another quick calibration test this evening when the stupid lousy sun goes away.

We're going to schedule another test in the near future in a different (bigger, darker) room.

The MoCap subject (i.e. me) is going to have to wear more muted colours as the cameras picked up flashes of white from highlights on my tennis shoes, my under-shirt and even reflections off my bald head! ...oh the indignity.

A more muted neutral background would help - I'm tempted to investigate a large chroma-key backdrop if we can't find a blank enough wall.

Overall I was pretty impressed with what we could see of the software but we're obviously a long way from being able to rubber stamp its usefulness at this point, but it's far from discounted.

On the subject of costs...

The cameras were approx £25 each (Amazon.co.uk)

I needed "Active" USB cable extensions (x4) for them as the cables they come with are pitifully short, that came to about £100 altogether (from Maplins)

The software has a 1 month trial period with full functionality in that time, if we decide to purchase it I'm looking at another $995 ...about £650?

You'll also need a USB hub - better be powered as well to cope with the active leads and cameras.

For a full days "proper" motion capture we're going to need a fairly large venue (as we don't have offices yet ...last of the bedroom developers) - I've investigated the local village hall (yes really, how cool is that!) and we're looking at *at least* another £125 there.

All in all ...we're staring about £1000 in the face to get up and running... but it isn't $10,000 a day.

It wasn't a wasted day, we set this up today to learn stuff and learn stuff we did.
We go into our next test better armed and I'd call that a win.

We'll get there in the end ...we always do.

I'll update again when we have more results, hopefully this is useful in some regard in the meantime.