The Kinect camera is built first and foremost to be a three-dimensional sensor, not so much a “camera” in regards to fidelity of the visible image. But what if that data could meet the optics filmmakers want, in a single, calibrated image?
The free and open source RGB+D Toolkit answers that need. It’s a combination of tools and knowledge – the mount to combine a Microsoft Kinect camera with your nice, DSLR camera, plus software for bringing everything together. But seeing it in action is the real joy. Check out the workshop results above, filmed at Barcelona’s HANGER, and overview below. Why do I have a feeling I’m going to be seeing this in dozens of music videos? (Very well; the tools are out there – now it’s up to you to make them compellingly unique in how you use them.)
The results: beautiful HD video footage from the camera, and loads of data in the Kinect image. From here, any number of data-driven effects – including a range of match-moved 3D – become possible. Consider these visuals just the beginning.
The Verge, which I can’t say enough is my favorite general tech blog, picks up on this in a story by Evan Rodgers. Kudos to them for noting innovations like this and not only every little gadget revision from big makers.
I notice in comments some common misunderstandings. Most notably, there isn’t just a choice between the Adafruit-led “hacked” software tools and Microsoft’s Windows-only tools; indeed, the most important option at the moment remains the OpenNI toolchain, including both open source and proprietary components. Those tools are cross-platform, and support camera hardware other than the Kinect hardware. (In fact, I could have rewritten this headline OpenNI, but let’s wait a bit so people know what I’m talking about.) My sense is that initiatives like this in the long run will gravitate around the multi-vendor, multi-OS OpenNI stuff, which in turn is supported by PrimeSense, who created a lot of the tech that powers the Kinect camera. We obviously need to do a proper post on the topic; if anyone would like to contribute research or input, let me know.
In the meantime, enjoy shooting some beautiful HD footage and do share how it goes.
And this comment, found on Tumblr from a Cnet story, makes me laugh.
Looks like a mess. I see no benefit to it. Why would a DSLR need to have a depth sensor? It actually already can manipulate depth, it’s called an aperture. This doesn’t make the video 3D or anything. You can already get a video to look like this out of the kinect – perhaps not as high quality, but this isn’t really high quality either…It’s all broken up and messy looking. Just some kids messing around. There’s really no application, you could throw a wireframe grid around objects in post production and various special effects with more accuracy.
Damn kids. Actually, what makes this comment interesting is that the commenter is responding entirely to the aesthetic look of the results. They’re looking at data, in other words, rather than an effect and failing to appreciate the difference. Aesthetically, their argument is entirely fair. But this reveals that people who haven’t seen raw visualized data may not understand the gap between that data and how it could be used. It’d be like looking at a construction site, seeing a bunch of bricks and mud around and tarps covering drywall, and saying:
You call this architecture? It’s just a big, dirty, ugly hole. Who would want to live there?
Of course, you could read my argument here as a bit of an indictment for those of us who decide to be satisfied with those raw building blocks and don’t take it further. Then again, if you’re happy…