Pitch — xtan.ai

AI Gesture Control for 3D Workflows ⚡️

💡 Idea

MotionCoder is a AI-powered framework that translates gestures and signs in real time into precise CAD/DCC commands — for faster 3D workflows, fewer clicks, and better ergonomics.

🚧 The Problem

Today’s 3D workflows (CAD/DCC) are click-heavy, time-consuming, and mode-bound: constantly switching modes, opening dialogs, typing parameters. That costs time, focus, and patience.

✅ The Solution

MotionCoder recognizes iconic gestures (e.g., line, circle, scissors, handwheel) and turns them in real time into commands with parameters.

CAD: Sketch → Extrude 12 mm → Chamfer 45° — in one flow, with one gesture.
DCC: Fast viewport navigation, cut/boolean/deform, blocking with fine-grained controls (handwheel gesture).

Designed primarily for the desktop with a large monitor — more comfortable & efficient than VR. VR is optional.

Why now? Modern GPUs with AI acceleration enable low-latency on-prem pipelines with high performance and strong cost-performance — often cheaper than ToF setups.

🚨 Gap

Similar ideas for gesture control exist, but they usually lacked end-to-end real-time capability — and the integration of sign-language nuances (handshape, fine-tuning, cognitive reference, spatial grammar, coarticulation, timing/rhythm, prosody, etc.). As a result, they saw little adoption or disappeared again. My entrepreneurial experience in sign language, programming, and mechanical engineering closes this gap.

⚙️ How it Works

Multi-view cameras detect gestures and reconstruct them as 3D gestures.
Real-time pipeline: 3D gestures → intent + parameters + continuous values (e.g., angle/Ø/depth).
Plugins control the target software (start: Blender, then Unreal, etc.).

🎯 Why It’s Compelling

Work faster: Fewer mode switches, more actions per second.
Ergonomic: Less mouse micro-work, more focus in the viewport.
Intuitive: Iconic gestures are self-explanatory.
Accessibility: Signs as a first-class interface ♿️
Secure & private: On-prem, no cloud required, data minimization.

👥 Who Benefits?

CAD teams: Set parameters faster, speed up reviews, keep focus.
DCC studios: Snappy navigation, non-destructive operators, live previz — creator speed.
Partners/integrations: Unreal, Blender, Autodesk, Dassault, Adobe (+ more).
Outside CAD/DCC: Sign-language assistance, medicine, assistive spatial interfaces, mechanical engineering/robotics (+ more).

⚔️ Competitive Landscape

Direct competition: No solution combines precise multi-view tracking, real-time semantics, and parametric CAD/DCC automation into one end-to-end pipeline.
Indirect/partial: Tracking stacks for VR/animation (e.g., MediaPipe, YOLO, Meta Quest) and standard hardware do not deliver industry-grade robustness for parametric CAD/DCC automation in practice.
Differentiation: MotionCoder closes the gap: intent + parameters in real time, undo-safe integrations into CAD/DCC.

👨‍🔧 Why Me?

Own CNC workshop; experience in mechanical engineering, drive technology, CAD, kinematics, C/C++ as well as sign-language communication. The idea for MotionCoder grows out of my CNC software development; plus hands-on vision/camera projects for clients.

From practice I know when MotionCoder is faster and more ergonomic than the mouse in everyday work. Gestures are more natural and faster in many scenarios — that shop-floor know-how is embedded in xtan.ai.

♻️ Conclusion

Gestures in → result out: fast ⚡️, precise 🎯, performant 💪.

More info: xtan.ai · GitHub: xtanai/overview