This paper presents Speech Playground, an interactive speech visualization and comparison tool designed to address the difficulties of integrating existing tools like Praat with modern deep learning representations.
The system combines a Python backend with a web-based frontend to enable the interactive exploration of multiple feature types, including continuous, discrete, and variable-length representations. It includes TextGrid and forced alignment support along with configurable distance and alignment settings for visual and auditory comparison.
Speech Playground is intended for use in speech research, representation validation, and computer-aided pronunciation training (CAPT)-oriented experimentation.