LlmHelper Audio

An MCP server for Claude Code that turns a script into voice clips with OpenAI text-to-speech: one numbered audio file per line (001.wav, 002.wav, ...) plus a manifest of each clip's duration, ready to lay onto a video. Works with Jig's demo_narrate to narrate demo recordings.

Before you start

Install LlmHelper and save your OpenAI API key in it (API Keys). The audio server reads the key from there; there is nothing else to configure.

Install

Run in PowerShell:

irm https://shakytext.com/audio/install.ps1 | iex

That downloads the server, checks its checksum, installs it to %USERPROFILE%\LlmHelper\mcp\audio and registers it with Claude Code as llmhelper-audio when the claude CLI is present. Re-run the same line any time to update.

To register by hand:

claude mcp add --scope user llmhelper-audio -- "%USERPROFILE%\LlmHelper\mcp\audio\LlmHelper.Audio.Mcp.exe"

Or grab the files directly: LlmHelper-Audio-mcp.zip (SHA-256) · README

Narrating a Jig demo

Ask Claude to: save a project with one segment per narration line, generate it (it asks for a spending limit; a 2-minute demo is about 3 cents), and export clips. Then film with Jig, putting a say mark with each line before its step, and run demo_narrate on the clips folder. Jig needs ffmpeg (it finds ShareX's copy or one on PATH). The README has the exact steps.

Requires the .NET 8 Desktop Runtime (the installer offers it via winget). OpenAI text-to-speech is billed to your own OpenAI key; the server refuses any call over the limit you give it and caps itself at $2 a day.