AI voiceovers in Claude and ChatGPT

Yes. Connect Makemancer and Claude, ChatGPT or Codex can turn your script into a voiceover with ElevenLabs v4 (the default), ElevenLabs Multilingual v2, Gemini 3.8 Flash TTS, MiniMax Speech 2.8 HD, Fish Audio Drama 3 or Kokoro. Paste the text, describe the voice, and get an audio file. 1,000 characters with ElevenLabs v4 cost $0.20.

Bring
Your script and a voice
Get
An MP3 of the narration
From
4 credits · $0.04

Made with it

Our own results, first takes. Each caption says what it took.

One script, three takes

The Oliva ad line from the second step below, read by ElevenLabs v4 voice Lily. The second take follows one note, written into the script: an ellipsis for the pause before “Oliva” and a [softly] tag on the last line. The third reads the same directed script in George, a deeper voice, to compare.

ResultA 7-second voiceover for the Oliva olive oil ad, ElevenLabs v4 voice LilyPressed by hand from olives grown on one hillside. Oliva. Taste the place it comes from.ElevenLabs v4 · 2 credits
ResultThe same voice after one note: a pause before “Oliva” and a softer last linePressed by hand, from olives grown on one hillside... Oliva. [softly] Taste the place it comes from.ElevenLabs v4 · 2 credits
ResultThe directed script in a deeper male voice, ElevenLabs v4 voice GeorgePressed by hand, from olives grown on one hillside... Oliva. [softly] Taste the place it comes from.ElevenLabs v4 · 2 credits

The same line on three models

One sentence about a corner bakery, read by ElevenLabs v4, Gemini 3.8 Flash TTS and MiniMax Speech 2.8 HD. Same words, different timbre, pace and price: hear them before you pick a model for a long script.

ResultThe bakery line read by ElevenLabs v4Every morning the bakery on the corner opens at six. By seven, the line reaches the street, and nobody minds the wait.ElevenLabs v4 · 3 credits
ResultThe bakery line read by Gemini TTS (Kore)Every morning the bakery on the corner opens at six. By seven, the line reaches the street, and nobody minds the wait.Gemini 3.8 Flash TTS · 3 credits
ResultThe bakery line read by MiniMax Speech HD (Deep Voice Man)Every morning the bakery on the corner opens at six. By seven, the line reaches the street, and nobody minds the wait.MiniMax Speech 2.8 HD · 3 credits

How to do it

  1. Connect Makemancer

    Add Makemancer to Claude, ChatGPT, Claude Code or Codex. The steps for each client are on the connection page.

  2. Give the script and the voice

    Paste the exact words to be spoken. Describe the voice (age, accent, pace, mood) or ask the assistant to list stock voices that fit before it generates. Most stock voices have a sample to play in the voice library.

    Read this script as a warm, unhurried voiceover for a 20-second ad, in a calm female voice with a light British accent: "Pressed by hand from olives grown on one hillside. Oliva. Taste the place it comes from." Use ElevenLabs v4 and tell me the cost first.
  3. Direct the next take

    Listen once, then ask for one change: a pause, more weight on a word, a softer line, or a different voice. The assistant writes the note into the script as punctuation or an audio tag. Only the new take is charged.

    Good. Add a short pause before "Oliva" and make the last line softer. Then read that version again in a deeper male voice so I can compare the two.

Which model to use

ModelBest forPriced forCredits
ElevenLabs v4speech/elevenlabs-v4The default: stock voices, expressive delivery with audio tags, 90+ languages.per 1,000 characters20$0.20
ElevenLabs Multilingual v2speech/elevenlabs-v2Stable professional narration in 29 languages.per 1,000 characters20$0.20
Gemini 3.8 Flash TTSspeech/gemini-3.8-flash-ttsStyle control and two-speaker dialogue from an exact transcript.per 1,000 characters18$0.18
MiniMax Speech 2.8 HDspeech/minimax-hd300+ voices with emotion, speed, pitch and volume control, 40 languages.per 1,000 characters17$0.17
Kokorospeech/kokoroFast, clean American English, good for quick drafts.per 1,000 characters4$0.04
Fish Audio Drama 3speech/fish-drama-3Multi-speaker dialogue with performance directions, and the voice for cloned or designed voices.per 1,000 UTF-8 bytes10$0.10

Tips for natural delivery

  • Write for the ear. Short sentences, numbers spelled the way they are said, and commas where you want a breath.
  • Describe the delivery, not just the voice. "Unhurried, smiling, like talking to one person" changes more than "nice voice".
  • Use audio tags with ElevenLabs v4. Square-bracket tags such as [whispers], [softly] or [laughs] set delivery inside the script. ElevenLabs Multilingual v2 does not read them.
  • Keep the voice for a series. Note the voice name from the first take and reuse it for every episode or ad, or save it on a Character.

Pick the model by the job

ElevenLabs v4 covers most narration and ads. ElevenLabs Multilingual v2 is the steady choice for long, plain narration. Gemini 3.8 Flash TTS and Fish Audio Drama 3 handle conversations between speakers. MiniMax Speech 2.8 HD gives fine control over emotion, speed and pitch. Kokoro is the quick option for American English drafts. For a voice that is yours, clone it and it speaks with Fish Audio Drama 3.

Put it under a picture

A voiceover can drive a video too. Give a portrait and the narration, and a talking avatar speaks it with matching lips; give an existing clip and lip sync moves the mouth to the new words.

Add music and sound

A voiceover often needs a bed underneath. Ask for a short music track, or a sound effect such as a door or rain (from $0.02), then ask the assistant to mix them with your narration or put them under a video.

Questions

How do I make a voiceover in Claude or ChatGPT?

Connect Makemancer, paste your script and describe the voice. The assistant picks a voice, tells you the cost, and returns an audio file with a permanent link. Without Makemancer, Claude has no text-to-speech of its own.

Can I use ElevenLabs in ChatGPT?

Yes, through Makemancer, if your ChatGPT plan and workspace allow custom MCP servers. You don't need an ElevenLabs account or API key.

Which voice model should I use?

ElevenLabs v4 for most ads and narration, ElevenLabs Multilingual v2 for long plain narration, Gemini 3.8 Flash TTS or Fish Audio Drama 3 for dialogue, MiniMax Speech 2.8 HD for fine control of emotion and speed, and Kokoro for quick drafts.

Can I control tone and pace?

Yes. Describe the delivery, add punctuation where you want a pause, and with ElevenLabs v4 put audio tags such as [softly] or [excited] in the script. MiniMax Speech 2.8 HD also takes exact speed and pitch settings.

Can it make a dialogue with two speakers?

Yes. Use Gemini 3.8 Flash TTS for two speakers or Fish Audio Drama 3 for several, and mark who says each line.

Can the voiceover be in my own voice?

Yes. Clone your voice once from a short recording of yourself, and every script after that can be read in it with Fish Audio Drama 3.

What does a minute of narration cost?

A minute of speech is roughly 800 to 1,000 characters. 1,000 characters with ElevenLabs v4 cost $0.20, and the cheapest model starts at $0.03; the assistant quotes before it runs.