A short and simple text carrying no emotion can represent some strong emotions when reading along with its context, i.e., the same sentence can express extreme anger as well as happiness depending on its context. In this paper, we propose a Contextual Affect Detection (CAD) framework which learns the inter-dependence of words in a sentence, and at the same time the inter-dependence of sentences in a dialogue. Our proposed CAD framework is based on a Gated Recurrent Unit (GRU), which is further assisted by contextual word embeddings and other diverse hand-crafted feature sets. Evaluation and analysis suggest that our model outperforms the state-of-the-art methods by 5.49% and 9.14% on Friends and EmotionPush dataset, respectively.