How to Generate Voiceovers From Claude With the Audexum MCP

AudexumOct 5, 20265 minBy Petar

Connect Audexum to Claude, ChatGPT, Claude Code or Cursor in one minute, then ask for voiceovers, narration and transcripts in plain words. Setup for each app, example prompts, engines and what it costs.

Claude can write your script. With one connector it can also read it aloud. Audexum's MCP server gives Claude (and ChatGPT, Cursor or any other MCP client) a voice: ask for a voiceover and you get back a link to an MP3. It can also transcribe recordings and pull YouTube transcripts.

There is nothing to install and no API key to copy. You paste one URL, sign in to Audexum, and click Approve.

Connect it in one minute

The server URL is the same for every app:

text

https://audexum.com/mcp

Claude (claude.ai or Claude Desktop)

Open Settings → Connectors → Add custom connector.
Name it Audexum and paste the URL above.
Click Connect, sign in to Audexum (or create a free account), and click Approve.

ChatGPT

Open Settings → Apps & Connectors → Create, paste the URL and choose OAuth.

Claude Code

Run this once in a terminal, or ask Claude Code to set up Audexum and it will run it for you:

bash

claude mcp add --transport http audexum https://audexum.com/mcp

Cursor, VS Code, Gemini CLI and others

Every client that speaks MCP over HTTP works. The setup page has a one-click Cursor link and a "Copy setup prompt" button you can paste into any assistant.

What to ask for

You talk to it in plain words. A few prompts that work:

You say	You get
"Make a 20-second warm, upbeat voiceover for my product launch."	An MP3 link, with Claude writing the script too
"Read this paragraph aloud in German."	Speech in any of 30+ languages
"Say it like an excited sports commentator."	The same text with a different delivery
"Transcribe https://example.com/interview.mp3"	The text of the recording
"Get the transcript of this YouTube video and summarise it."	A summary built from the real transcript
"How many Audexum credits do I have left?"	Your plan and balance

Four voice engines

Claude picks a sensible engine by itself, but you can name one. The best one is V4, which also performs acting tags you put in the text, such as <laugh>, <sigh> or <whispers>.

Engine	Best for	Plan
V4 Real	Voiceovers, ads, podcasts, characters. Style direction and acting tags	Creator plan and above
V3 Expressive	Natural named voices for English narration	Every plan, free included
V2 Multilingual	31 languages, including Bulgarian, Greek and the Nordic and Baltic languages	Every plan, free included
V1 Instant	Long English drafts where cost matters most	Every plan, free included

What it costs

A free account gets 2,000 credits at signup and 250 more every day. One credit is one character times the engine's multiplier (V1 1x, V2 1.5x, V3 2x, V4 4x), so a 60-second voiceover on V3 costs about 1,800 credits. YouTube transcripts are free. The Creator plan (EUR 12 a month) adds 480,000 credits and V4. See pricing for the other plans and credit packs.

Privacy

Audexum is hosted in the EU. The connector only sees what Claude sends it for the task you asked for, and audio links expire after 30 days. Section 3 of the privacy policy covers AI assistants connected through MCP.

Ready? Open the setup page and connect it in one minute.

Got any questions left?

What is the Audexum MCP server?How do I get an API key and make my first call?How do I make my first voiceover?

Share feedback

Was this article helpful?

Next articleBest Windows Dictation App 2026: 5 compared on price, platforms and what else you get