Conditional diffusion models play a significant role in generative AI tools like ChatGPT, DallE, and Sora. At fist glance these applications appear to be very different from probabilistic inverse problems, which is the topic of this manuscript. However, the core feature of these algorithms is to use samples from the joint density of observations and quantities of interest to create a tool that can generate samples from the conditional density of quantities of interest conditioned on a given observation. In this manuscript, we demonstrate that a tool that accomplishes this can be easily adapted to solve inverse problems. Thereafter we apply this tool to solve inverse problems motivated by applications in mechanics. In addition, the “traditional” derivation of diffusion models is rooted in principles of stochastic differential equations and is therefore not accessible to readers who are not versed in this topic. In this manuscript, we provide an alternative derivation that utilizes probability densities rather than samples or particles, and uses elementary knowledge of partial differential equations.