mcps
E

Recipe

Turn your words into a real voiceover

Paste your script, direct the delivery — tone, pace, pauses — and get a natural voiceover file back. The direction is the skill, not the script.

for Anyone who wants a short video from a photo or a sentence — no editing software

Hand-checked recipe · curated by Ilya Gindin

10 minutes

to run

beginner

difficulty

1

connector

What you'll need

ElevenLabs Turns your text into a lifelike voice and saves the audio.
Required

This recipe runs in 3 steps

  1. 1

    Direct the read, not just the text

    Say to your AI

    Read this in a warm, conversational voice with natural pauses — relaxed, not announcer-style: [paste your script].

    You'll get: A human-sounding take. The tone words are the lever: without them the default read is flat.

  2. 2

    Fix names and acronyms

    Say to your AI

    Same read, but pronounce 'mcpservers' as 'M-C-P servers' and keep the brand name exactly as written.

    You'll get: Correct pronunciation — spell tricky words phonetically and it stops guessing.

  3. 3

    Match the voice to the job

    Say to your AI

    Now try it as a calm documentary narrator, slightly slower — and export the better take as MP3.

    You'll get: Two directions to choose from, and a ready-to-use file.

You're done when

A natural-sounding voiceover from your script in minutes — no microphone, no recording, and a delivery you directed instead of defaulted into.

One connector, ten minutes: paste the script, direct the delivery, export the take. The whole trick is that the words you say ABOUT the read matter more than the script itself — «warm, conversational, natural pauses, not announcer-style» is what turns text-to-speech into a voiceover.

What the README won’t tell you: the default read is flat by design; tone direction is the lever. Unusual names get guessed unless you spell them phonetically. And long scripts work better fed in sections with the same direction repeated — the voice stays consistent and the quota stretches. Pair the result with the talking-head recipe and the captions come out matching the voice for free.

Checked 2026-07-12: ElevenLabs official connector live (uvx install + API key; free tier 10k credits/month).

If something breaks

I can't find ElevenLabs to turn on
Fix: Official connector: install with uvx elevenlabs-mcp and your ELEVENLABS_API_KEY (free tier has 10k credits/month), or add it from your AI's connector settings.
The voice sounds flat or robotic
Fix: Direct it: 'warm, conversational, natural pauses, not announcer-style'. Delivery words change the read far more than the voice choice.
It mangles a name or acronym
Fix: Spell it phonetically in the prompt ('GIF as jif', 'M-C-P servers'). It guesses on unusual words unless told.
Long script cuts off or drifts
Fix: Feed long text in sections and keep the same delivery direction on each — it reads chunks better than walls, and your character quota stretches further.

People ask their AI

“make a voiceover from text”“read my script in a real voice”“add narration to my video”

Related

ElevenLabs talking head explainer from a script