Live Translate Mcp

dev.waxberryPublished by waxberry.devv0.1.6Updated Oct 24, 2026

MCP server for local speech translation (EN ↔ 中文) via Whisper + Claude + Piper

VerifiedSTDIODesktop onlyMedia & DesignAI & ML

Overview

AI-generated overview

Local MCP server that lets Claude transcribe, translate and speak English and Mandarin audio using Whisper, the Claude API and Piper TTS.

What you can do
Translate an audio file: transcribe a WAV file, translate it via Claude, synthesise speech, save the translated WAV next to the original and play it, returning the original text, translation and output path. Translate raw audio: accept base64-encoded WAV and return transcription, translation and synthesised audio as base64. Check readiness: verify the Whisper model cache, Piper voice files and espeak-ng are present before translating. Language is detected automatically between English and Mandarin.
When to use it
Useful when you want Claude to handle English–Mandarin speech translation from audio files on your own machine, with audio kept local and only the translated text sent to the Claude API. Suited to meeting clips, interviews or recordings you want transcribed, translated and spoken back.
Requirements
Runs locally via npx or a global npm install; needs Node.js 18+, an Anthropic API key in ANTHROPIC_API_KEY, and espeak-ng installed. The Whisper model (~150 MB) and Piper voices (~200 MB) download on first use and are cached in the home directory; PIPER_VOICE_DIR can override the voice path. Desktop only.
Before you install
Requires an Anthropic API key, which is a billable credential; translation calls incur Claude API costs. Translated text is sent to the Anthropic API, though audio stays on device. The server writes a translated WAV file next to the source and plays audio automatically. First run downloads several hundred MB of models.

Installation

In SourceWeft

  1. Open Live Translate Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

[Waxberry]

live-translate-mcp

[live-translate-mcp MCP server]

Real-time English ↔ Mandarin speech translation for Claude — powered by Whisper, Claude AI, and Piper TTS.

Give Claude the ability to listen, translate, and speak. live-translate-mcp is a Model Context Protocol (MCP) server that adds speech translation as a native tool inside Claude Desktop and Claude Code. Hand it an audio file, and it transcribes, translates, synthesises, and plays the result — entirely on your machine, with Claude handling the translation.

[live-translate-mcp demo]

What it does

StepTechnologyWhere it runs
Speech → TextOpenAI Whisper (via @huggingface/transformers)Local
Text → TranslationClaude (Opus 4.8)Anthropic API
Translation → SpeechPiper TTS (ONNX)Local

Audio never leaves your machine except for the translated text sent to the Claude API. ASR and TTS run fully on-device.

Tools

translate_file

Translate a WAV audio file. Pass an absolute path — the server transcribes it, translates the text via Claude, synthesises speech, saves <name>_translated.wav next to the original, and plays it automatically.

Translate /Users/alice/meeting_clip.wav

Returns: original text, translation, and the path to the saved output file.

translate_speech

Translate raw audio passed as a base64-encoded WAV string. Returns the transcription, translation, and synthesised audio as base64 WAV — useful for programmatic workflows.

health_check

Verify that all dependencies (Whisper model cache, Piper voice files, espeak-ng) are present and ready before making a translation request.

Installation

No installation required. Run it directly with npx:

bash
npx -y live-translate-mcp

Or install globally:

bash
npm install -g live-translate-mcp

Prerequisites:

  • Node.js 18+
  • An Anthropic API key
  • espeak-ng — for TTS phonemisation (brew install espeak-ng on macOS, apt install espeak-ng on Linux)

The Whisper model (~150 MB) and Piper voice models (~200 MB) download automatically on first use and are cached in ~/.live-translate/.

Claude Desktop setup

Add the following to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

json
{  "mcpServers": {    "live-translate": {      "command": "npx",      "args": ["-y", "live-translate-mcp"],      "env": {        "ANTHROPIC_API_KEY": "your-api-key-here"      }    }  }}

Restart Claude Desktop. You'll see live-translate appear in the tools panel.

Claude Code setup

bash
claude mcp add live-translate -- npx -y live-translate-mcp

Then set your API key in the environment before starting Claude Code, or pass it via the MCP env config.

Usage examples

Once configured, just ask Claude naturally:

"Translate this audio file for me: /Users/alice/recording.wav"

"Use translate_file on /tmp/interview.wav"

"Check if live-translate is ready"

Claude will call the appropriate tool automatically.

Supported languages

LanguageASRTranslationTTS
English✓✓✓
Mandarin Chinese (普通话)✓✓✓

Language is detected automatically from the audio — no need to specify it.

How it compares

live-translate-mcpCloud speech APIs
ASRLocal (Whisper)Remote
TranslationClaude APIRemote
TTSLocal (Piper)Remote
Audio privacyAudio stays on deviceAudio uploaded
CostClaude API onlyPer-minute pricing
OfflinePartially (ASR + TTS)No

Environment variables

VariableDefaultDescription
ANTHROPIC_API_KEY—Required. Your Anthropic API key.
PIPER_VOICE_DIR~/.live-translate/voicesPath to Piper .onnx voice files.

Related

  • live-translate — the standalone CLI for push-to-talk translation with no API key required, using local models end-to-end.
  • Model Context Protocol — the open standard this server implements.
  • Piper TTS — the local TTS engine powering speech synthesis.

Licence

MIT

Source: README.md at commit 8d4a4af

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.6LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts