Conventional secret message passing methods embed a message in the cover-text, so the receiver must use stego-text to extract the message content. In contrast, this paper proposes a new paradigm in which the receiver does not necessarily require stego-text to retrieve the message content. Under the proposed approach, the sender can produce keys without modifying the cover-image, and the intended recipient can use the keys and an image that resembles like the cover-image to recover the message. The feature has the potential to generate many new secret message passing applications that would probably be impossible under the current widely used paradigm. The performance criteria of the new paradigm are presented. We propose a subspace approach to implement the paradigm, demonstrate that the criteria can be satisfied, and consider some interesting application scenarios.