Designing User Interfaces (UIs) that can interpret and respond to user emotions has evolved from a passing trend to a core tenet of design philosophy. As technology advances, it becomes increasingly important to predict users’ emotional states during interface interactions. This paper proposes a cutting-edge deep learning model designed to predict a wide range of emotions in users, such as Happiness, Anger, Calmness, and Surprise. The research started with a thorough survey, gathering feedback from 72 users regarding UI designs tailored to various emotions. Using these insights, interfaces were developed to address different emotional states. Following the design phase, a user interaction study was conducted involving 29 participants interacting with these interfaces while their facial expressions were recorded. Subsequently, a predictive model for user emotions was developed, achieving an accuracy of 83%, showcasing its robustness and reliability in discerning and predicting user emotions based on interactions with UIs. This model was seamlessly integrated into a proposed real-time adaptive UI system. Notably, existing literature has traditionally relied on pre-existing software for emotion detection in adaptive systems. Our contribution is to demonstrate an Emotional Interface Adaptation System that uses a trained model in real-world circumstances to provide a dynamic and responsive user experience that adapts in real-time to users’ emotional cues.