Hand gesture recognition is turning into a extra prevalent mode of human-computer interplay, particularly as cameras proliferate throughout on a regular basis units. Regardless of continued progress on this subject, gesture customization is usually underexplored. Customization is essential because it permits customers to outline and exhibit gestures which might be extra pure, memorable, and accessible. Nevertheless, customization requires environment friendly utilization of user-provided knowledge. We introduce a technique that allows customers to simply design bespoke gestures with a monocular digital camera from one demonstration. We make use of transformers and meta-learning methods to deal with few-shot studying challenges. Not like prior work, our technique helps any mixture of one-handed, two-handed, static, and dynamic gestures, together with completely different viewpoints, and the flexibility to deal with irrelevant hand actions. We implement three real-world purposes utilizing our customization technique, conduct a consumer examine, and obtain as much as 94% common recognition accuracy from one demonstration. Our work offers a viable path for vision-based gesture customization, laying the inspiration for future developments on this area.