I have always been frustrated with the difficulty of posing figures. I know, I know...it's supposed to be easy (or at least manageable). If such were truly the case, then why the lucrative trade in "canned" poses?

I had an idea today that may be worth exploring. What if one took a little wooden, articulated human figure, embedded it with appropriate sensors, and use it to communicate 3D poses directly into Poser? I think it would be a joy to use such an interface, and would gladly pay big bucks to someone who could actually develop it.

Are you listening, propeller heads?

Hmm. The figure construction could be handled by just about any 3D printer. The issue would the number of sensors you would have to pack onto the thing to get the minimum number of data points. A Raspberry PI could handle the signal translation, but the most you could expect would be basic XYZ trans. Each value would add a sensor to the armature, or require more complex articulation. You could technically do it with digital gyro's like they use in small drones and get triple axis data, but that would still be a gyro for every joint, and the armature scale would increase as more sensors were added. It really isn't a case of can it be done; it already has been. But the woody model was so expensive that it wasn't commercially viable.

@Dale B Much obliged for your quick and very helpful post. Maybe more can be accomplished with less - there is no need for fine granularity in the captured geometry; a first-pass, rough pose might be good enough to fine-tune manually later.

Google gyroscope sensors and you will get an idea of prices. You have the cheap accelerometers like Adafruit for around $15 a sensor, but the sensitivity probably isn't sufficient. Those cheaper units have peak rates of 200 gravities of acceleration, and are used for rocketry and impact studies. The more flexible gyro's start at around $30 and go up towards $700 or so, depending on the sensitivity. So you would have to prototype just to see how sensitive your sensor has to be to even work. Just moving a leg or arm would involve only fractions of one g acceleration, so the sensor would need to be capable of handling such low force values reliably. Just getting basic body motion would take at least 17 sensors, allowing for ribcage and two points below that for spinal flex. Even using the cheap Adafruit sensor, that's $250 just for sensors. And that is only for 'Woody' style input; no hand motion except wrist bending, and the same with ankles. You would have to design the process controller scheme, as it would take more than one or two of them to handle the sensor input, write the software routines to poll the sensors, and then choose the interface method and program that. Even with a custom wiring harness, the figure armature would probably be at least 18-24" tall just to get everything on the armature. If you wanted it conceald within it, you'd be looking at around 30". There are other sensors, like the single plane flex tape that only measure action along one plane of motion, but then you get into the issues of mixing sensors, which complicates the software by another level. Like I said it can and has been done, but the cost would be significant, as in probably in excess of a grand to prototype.

There is no need for expensive sensors. Look at ebay. Chinese sellers asks 1-3 dollars for a simple gyro.

But I wouldn't use so mutch gyros. I would use variable-resistors (https://www.ebay.com/sch/i.html?_from=R40&_trksid=m570.l1313&_nkw=variable+resistors&_sacat=0)

Analog resitors are mutch better for this kind of task. 3 additional gyros (head, cheast, hip) and you will have anything you need.

Another, mutch simpler solution is doing it with 2-4 cheap cameras and colored leds on the figure (manequin).

By the way, instead of an expensive RasPi I would use cheap ESP32's. They are much better suited for something like this.

Or, for real flexibility, one ATTINY on each variable resitor, all connected via I2C to a central ESP32.

For the story with the cameras, there are also solutions based on ESP32. They cost about 10 Euros each (ESP32 microcontroller with cam on top). Together with a few LEDs and other parts, a manequin printed on the 3D printer costs less than 100 euros. On top of that, however, there are several hours of tinkering and programming.

For the sake of completeness: About half a year ago I started with a Python lib that connects Poser to MQTT. The currently selected figure in Poser is controllable by it.

But the lib is not quite finished yet. I used it for testing, and found out that the external control of Poser is nothing more than a nice gimmick.

For the simple hinge joints variable resistors would be sufficient. They would be next to useless at complex joints like hips and shoulders due to size constraints. Ribbon sensors could be used in them, but would require you to recreate the articulation of those joints. And there would be joint and linkage maintenance to consider.

I haven't seen those process controllers or dipped into the euro market lately, so I'll have to check them out.

Nice idea on the micro mo-cap armature. The software would be the kicker, though, and it would require enough space to lock the camera locations to prevent vector drift. Even a Barbie sized armature would need around a 15" cubed capture field, and nothing to distract the IR cameras.

And yeah, gimmick is about what it would be. A 3D mouse is a nice added controller, at least as long as the sensitvity is cranked way down.

https://www.youtube.com/watch?feature=player_embedded&v=MfJ4pA8ngDo

The Japanese already sell pre-made pose-to-software mannequins like this, such as Qumarion for use with the 3D figures in Clip Studio Paint. I guess someone could code a driver to make it work with Poser, if the Japanese haven't already done that? Or, for the wire-heads, a Qumarion might be reverse engineered for Poser, and a simple DIY kit produced at much less cost.

See also the Nov 2019 thread here titled: "Could USB Poseable Mannequins become useful?".

This is so funny, Poser was written to allow a digital mannequin, that would replace a physical one, to be posed so that it could be used as a reference, now it is suggested, quite reasonably, that we create a physical mannequin to pose as a reference for the digital, awesome :)

Yeah, I'd much prefer using something like Microsoft kinect if I'm going to go that route.

EClark1894 posted at 6:50AM Fri, 28 February 2020 - #4382022

Yeah, I'd much prefer using something like Microsoft kinect if I'm going to go that route.

Funny thing, but the Microsoft made something kinda specific for applications just like this: https://azure.microsoft.com/en-us/services/kinect-dk/

Oh, and they did most of the heavy lifting code-wise: https://github.com/Microsoft/Azure-Kinect-Sensor-SDK

...want it to work on Macs, you say? Okay: https://github.com/OpenKinect/

Looks fairly non-evil to do.