In this paper, we present a technique for the discovery of hybrid process models that combine imperative and declarative constructs. In particular, we first employ the popular Inductive Miner to generate a fully imperative model from a log. Like most imperative miners, the Inductive Miner tends to return so-called flower models for the less structured parts of the process. These parts are often imprecise. To counter these imprecise parts, we replace them with declarative models to increase the precision since declarative models are good at specifying which behavior is disallowed. The approach has been implemented in ProM and tested on several synthetic and real-life event logs. Our experiments show that hybrid models can be found to be more precise without overfitting the data.