Pronunciation & Voice Coach
Add local English pronunciation, grammar, and fluency coaching to Claude or Cursor via MCP without sending audio to a cloud API.
Overview
Pronunciation & Voice Coach is a build-phase MCP server that gives agents local English pronunciation, grammar, and fluency feedback using faster-whisper.
What is this MCP server?
- Runs locally with faster-whisper; model sizes from tiny.en through large-v3-turbo
- stdio MCP transport via uvx on PyPI (mcp-server-pronunciation 0.3.0b3)
- Optional HF_HUB_CACHE for Whisper weight storage
- Feedback on pronunciation, grammar, and fluency from captured or supplied audio
- Server schema version 0.3.0b3 on PyPI package mcp-server-pronunciation
- Default model base.en; options include tiny.en, small.en, medium.en, large-v3, large-v3-turbo
- Transport: stdio; registryType pypi
What problem does it solve?
You want speaking-practice or voice QA inside the agent loop but refuse to ship raw audio to cloud transcription vendors.
Who is it for?
Indie builders adding voice coaching to agents or improving spoken English while staying local-first.
Skip if: Teams that need non-English coaches, enterprise call-center QA, or hosted-only STT with no local GPU/CPU budget.
What do I get? / Deliverables
After registration, agents call MCP tools that analyze speech on your machine and return actionable language feedback.
- Registered stdio MCP server in agent config
- Pronunciation and grammar feedback tool calls from the agent
- Tunable faster-whisper model via environment variables
Recommended MCP Servers
Journey fit
Voice-coach MCP extends the agent stack during product build, not a launch or ops workflow. Fits agent-tooling: stdio PyPI package (uvx mcp-server-pronunciation) with faster-whisper models configured via MCP_PRONUNCIATION_MODEL.
How it compares
Local speech MCP integration, not a browser skill or generic ChatGPT voice mode.
Common Questions / FAQ
Who is Pronunciation & Voice Coach for?
Solo and indie developers who use Claude Code, Cursor, or similar agents and want on-device English pronunciation and fluency coaching over MCP.
When should I use Pronunciation & Voice Coach?
When you are building or testing voice-aware agent flows and need grammar and pronunciation feedback without a cloud speech API.
How do I add Pronunciation & Voice Coach to my agent?
Add the io.github.JuhongPark/pronunciation MCP entry, run via uvx mcp-server-pronunciation (PyPI), and set MCP_PRONUNCIATION_MODEL or HF_HUB_CACHE if needed.