Since the concept of information hiding without covering was proposed, it has seen significant development due to its effectiveness in resisting steganographic tools and its robustness to degraded container images. However, existing methods do not perform well in controllability, security and robustness. To address these limitations, inspired by recent developments in diffusion models, we propose Okey, a powerful image steganography method with the following characteristics: (1) it can convert any secret image into a hidden image with enhanced control; (2) it enhances security through non-natural language private key and makes hidden images difficult for humans to perceive; (3) it robustly recovers the secret image using the reversible conversion capability of the diffusion model. Specifically, our proposed method is based on optimized key, using the null-text embedding condition of the diffusion model as a trainable key. To sum up, we introduce a groundbreaking approach to image steganography called Optimized Key (Okey), which offers significant advantages in controllability, security and robustness compared to previous state-of-the-art image steganography methods. We conduct detailed experiments to demonstrate the advantages of the proposed Okey.