A DIY Robotic Cameraman Built From Salvaged Electronics

Robotics isn’t the easiest field to get into by a long shot. There is a lot to learn in areas ranging from computer science and mechanical engineering to artificial intelligence (AI), but that’s not what holds most people back. Rather, it’s the high cost of the hardware that is the biggest hurdle. However, if you are willing to put in the time and effort, building robots from scratch can save you a lot of money.

These DIY robots can also be very capable, as YouTuber WhippetInTech’s latest creation demonstrates. He has been working on a robotic arm recently, and just finished up a round of updates that turned it into an autonomous, AI-driven camera operator. With a bit of elbow grease and a number of wise design choices, WhippetInTech was able to keep the robot highly accessible.

The robotic arm was originally assembled largely from salvaged 3D printer parts, but the system had a flaw in its shoulder joint that caused mechanical backlash. The arm relied on a 100:1 planetary gearbox that introduced just enough play to make the entire structure unstable. Because the arm was fairly long, even tiny amounts of slack became dramatically amplified, causing the camera mount to sag and wobble whenever the arm changed direction.

To fix this issue, WhippetInTech upgraded the shoulder assembly with an inexpensive harmonic drive sourced from AliExpress. Harmonic drives are highly prized in robotics because they can provide large gear reductions while eliminating nearly all backlash.

Integrating the new gearbox into the arm was not straightforward, however. Since the harmonic drive lacked a traditional output shaft, a completely custom coupling system had to be designed in CAD. Early attempts to manufacture the parts through aluminum casting failed because the tolerances were too loose, ultimately forcing a switch to CNC machining and lathe finishing. After several iterations, the final metal coupling achieved the rigidity and precision needed for smooth movement.

Once the mechanical upgrades were complete, the robot underwent calibration and stress testing. The arm uses a Field Oriented Control motor driver that requires tuning to accurately map motor positions to the new drivetrain. Payload tests showed that the upgraded shoulder performed reliably with loads up to roughly 1.25 kilograms. Beyond that point, the elbow stepper motor began skipping steps.

This weight limitation heavily influenced camera selection. Instead of using a large mirrorless camera, the creator chose a lightweight Sony Cyber-shot DSC-QX10 lens-style camera weighing only about 130 grams. To design a perfectly fitted mounting cradle, the camera was scanned using a photogrammetry workflow involving matte spray coating, dozens of reference photos, and reconstruction inside Artec Studio.

The robot’s autonomous tracking system relies on a second camera entirely dedicated to AI vision. A lightweight USB webcam streams video to a Mac Mini running YOLO object-detection software. The custom tracking software calculates the center of the detected subject and continuously commands the robot’s pan and tracking axes to keep the person framed in the shot.

To complete the build, the creator fabricated a futuristic camera housing using both resin and FDM 3D printing techniques. Brass heat-set inserts strengthened the printed parts, while an integrated ball bearing provided smooth movement for the camera head. An ESP32 microcontroller and glowing LED accents added a final sci-fi touch to the fully autonomous robotic cameraman.