VocalSync: AI-Based Speech Delivery Analysis System
Abstract
Good public speaking is one of the most valuable professional skills. However, many people cannot get personal speech coaching because it is alot of expensive and there are not enough qualified trainers available whcih can fulfill demand at large scale. We present VocalSync, a web-based platform that delivers objective of each person,data-driven analysis of spoken communication without requiring human instructors. it is built around a record-process-feedback loop, the system chains together three AI/ML modules: OpenAI Whisper handles speechtotext conversion, DistilBERT performs emotion and confidence classification, and the Groq LLM API synthesizes the results into tailored coaching reports. Beyond qualitative assessment, VocalSync quantifies delivery through words-per-minute (WPM) measurement and filler-word counting, and retains session histories so that speakers can monitor their own development over time. On the GoEmotions benchmark, the classification module reaches 82.4user study involving both professional coaches and graduate students produced a Mean Opinion Score (MOS) of 4.1 out of 5, with raters judging 78comparable to expert human feedback. Technically, VocalSync follows a hybrid edge-cloud model: a Next.js frontend handles recording and presentation, a FastAPI service manages inference, and a MySQL database stores longitudinal records. This work demonstrates that combining acoustic metrics, transformer-based emotion recognition, and LLM-driven coaching into a single pipeline can provide speech assessment quality that was previously available only through professional speech coaching.. Index Terms—Speech analysis, confidence assessment, natural language processing, emotion detection, large language models, DistilBERT, Whisper, public speaking, AI coaching, speech delivery metrics. Index Terms-Speech analysis, confidence assessment, natural language processing, emotion detection, large language models, DistilBERT, Whisper, public speaking, AI coaching, speech delivery metrics.