People are increasingly using social media platforms to express their feelings, thoughts, and opinions, producing a large amount of sentimental content every day. While customary sentiment analysis bases its data on text, images and audio come with contextual information. Thus, this integration has given birth to the method of “multimodal sentiment analysis,” where data coming from more than one source is compiled together to enable the appreciation of social media sentiment in its entirety. Multimodal sentiment analysis will capture subtle user feelings and sentiments through the combination of visual and auditory elements with the transcript. This work discusses issues and challenges related to MMSA for social media. This research, in general, scans through accessible prose sources on new strategies and models, like the (MTFN-CMM) and (UCRNN). Subsequently, this work will further explain the analysis of sentiment on social media, its boundaries, and applied multimodal sentiment analysis in the multidomain applications encompassing analyses of variety discernment, crisis management, and political sentiments. Apart from this, this work envisions challenges in scalability, isolation, and cross-modal demonstration learning while proposing the information toward more encroachment. Multimodal sentiment analysis is a new trend in academia and deals with providing a simple yet precise, granular sentiment analysis of this energetic world of social media.