People have the ability to manage and control their emotions. They frequently are aware of their feelings. Consider what may happen if someone is unable to identify their emotions. There is a chance that it is problem connected to mental illness, this may be the case. The first step is early symptom detection. The method to predict levels of depression and emotions is presented in this paper. Firstly, the analysis of a DAIC-WOZ dataset involves using several Deep Learning models with three modalities—text, audio, and video features—are combined to predict patient depression. This method of fusion regulates the amount of contribution from each modality. The candidate can evaluate their own mental health at the pre-diagnosis stage. To do this, text and voice data are combined to create a system that can estimate the severity of depression. Chats are the textual data as inputs. Voice recognition is performed on audio data, and the audio is then transformed to text representation. By classifying the emotions into several emotional levels—such as angry, sad, etc.—the emotions are recognized. The user learns about their degree of depression, which helps to some extent with the diagnosing portion of treatment.