YouTube has become a major platform for comment, review, and opinion-sharing, especially through comments. Understanding the emotions of such comments can be extremely helpful to content creators, advertisers, and researchers. In this research work, we present an automated sentiment analysis system that can analyze YouTube comments in both English and Bangla. Through the Google YouTube API, we retrieve comments through the snippet feature. Since the dataset has mixed-language input, we preprocess the comments initially by deleting stop words, punctuation, and emojis. NonEnglish comments, which are primarily in Bangla, are detected using the Lang detect library and translated to English to have a uniformity in the dataset. The data is additionally cleaned and tokenized post-translation. We utilize the Text Blob library to find out the polarity (positive, negative, neutral) and subjectivity of every comment. To facilitate better feature representation, TF-IDF vectorization with n-grams is done. For dealing with class imbalance, we have used the SMOTE approach. Different machine learning algorithms are trained and tested by hyperparameter tuning. Logistic Regression achieved highest accuracy of 95% on C=10, and following that was Support Vector Machine (SVM) with a linear kernel's 94%. Our final system accepts as input any YouTube video link and returns an analysis of public sentiment based on the comments. Beyond enabling multilingual sentiment analysis, this is also a simple way of tracking feedback in real time on a range of linguistic content.