Human beings express emotions and feelings in various ways, the intonation with which we express ourselves, our facial expressions or the way we write are key to detect how we feel in response to certain situations. Nowadays, the large amount of news published on the many social networks makes it impossible to check all this information manually, which is why systems have been created to detect “fake news” automatically. “Fake news” often use emotional language to capture the reader’s attention, so sentiment analysis, capable of identifying emotions in text, helps to detect them. The main objective of this project is to provide a service capable of multimodal analysis of audio, text and video to recognize the user’s feelings, emotions and moral values. Thanks to the development of this system it will be possible to combine the results of individually analyzing audio, text and video in previously created models to obtain a more accurate result. To achieve this development we will start by creating an API with “mock” requests where we will simulate the result of different requests both individually for each of the modes, audio, video, and text, as well as multimodal requests that aggregate the results obtained through individual requests. Subsequently, a multimodal dataset will be used and the “mock´´ requests will be replaced by real requests to the analysis services developed for each of the modalities. In order to evaluate the performance of the services, different forms of evaluation will be used to compare the expected result, the one present in the dataset, with the one obtained through the requests to the services. All this will be done using Python with several of its libraries and Fast API technology to create the API from which the requests to the different services are redirected.