Recycling is crucial for sustainability, yet people are often afraid to recycle due to the fear of recycling wrong. This research introduces AutoRecycle, an intelligent recycling machine designed to automate waste sorting using advanced machine learning techniques. Using the RealWaste dataset, we created an ensemble model that utilizes both a Vision Transformer (ViT) model trained with the Self-Supervised DINO method and a Swin Transformer. We also created a proof-of-concept physical machine that includes a high-resolution camera, a servo motor, and an ultrasonic sensor integrated with a Raspberry Pi for real-time image processing. Testing results demonstrate that our ensemble model achieved 95.49% test accuracy in detailed waste classification and 98.73% in binary recyclability classification. AutoRecycle proves to be a scalable prototype that can significantly improve effective recycling rates and thus promote more recycling.