Speech-To-Text (STT)/Transcription Services Benchmarking Tool

A cross-platform benchmarking tool for evaluating Speech-to-Text (STT) and Text-to-Speech (TTS) systems.

Project for evaluating and improving LLMs, featuring a tool for benchmarking models and a chat system with text + voice support. Deployed on PAL Robotics’ ARI robot with STT, TTS, and real-time audio, enabling natural human-like conversations.

A full-stack tool for developers to benchmark ASR models (Whisper, Google STT, etc.). The React/Vite frontend provides a UI to upload audio and a ground truth transcript. This is sent to a Python/FastAPI backend, which calculates and compares Word Error Rate (WER), Character Error Rate (CER), and processing latency for each model.