Back to projects

English Speak Reviewer

A microphone-first practice platform designed to help users improve spoken English through structured recording, analysis, and feedback.

Hono HTMX Python Whisper Wav2Vec2 BullMQ Redis MinIO Docker

Overview

English Speak Reviewer is a practice platform built to simulate interview-style speaking scenarios and provide structured feedback on spoken responses. Users record audio directly in a browser-based studio, after which the audio is stored in object storage. A separate Python worker consumes jobs from a Redis-backed queue to perform speech processing tasks such as transcription and alignment using Whisper-based models. The system then derives fluency-related metrics, such as speaking rate and filler usage, and optionally generates narrative feedback to help users understand areas for improvement.

Key Challenges

  • Designing a system where latency-sensitive user interactions remain responsive while speech processing tasks are computationally heavy.
  • Coordinating asynchronous job execution and tracking processing states across API and worker boundaries.
  • Managing storage and retrieval of large audio artifacts reliably in a containerized environment.

Outcomes

  • Delivered a stable practice workflow that provides structured, actionable feedback on spoken responses.
  • Successfully separated resource-intensive speech processing from the user-facing application layer.
  • Established a modular, container-based architecture that can be extended with additional analysis steps.

Interested in this project?

Let's discuss how we can build something similar for you.

Contact Me