How to Clone Your Voice and Use Text to Speech for Video Voiceovers

To clone your voice for video voiceovers, open Voice Studio, upload 10 to 30 seconds of yourself talking in a quiet room, name the voice and press Clone; after that you type any line, pick your voice, and the take lands in your Library as audio ready for a video. In ReelMint a clone costs 375 credits once and reads in 30+ languages including Arabic; if you do not need your own voice, the built-in text to speech voices are cheaper still.

We build ReelMint, an AI studio that generates videos, images and music and posts them to TikTok, YouTube, Instagram, Facebook and X. Here is exactly how the voice tools work, what they cost and where the audio goes.

Clone your voice, step by step

1. Open Voice Studio. It sits in the Studio menu after Characters. You can also reach it from any audio slot in Mint Studio through Speak in my voice.
2. Record a clean sample. 10 to 30 seconds of you speaking normally: no music, no echo, no second voice. A phone in a quiet room is enough. The clone is only ever as good as the sample.
3. Name it and clone. Upload the recording (audio or video files both work), name it ("My voice"), and press Clone this voice. It takes up to a minute and costs 375 credits, once.
4. Read a line. Pick the voice, type what it should say, and press Speak or Listen. A take is up to 10,000 characters and takes a few seconds. It is filed in your Library as an audio post.

The clone is permanent. It stays in your roster and is pickable everywhere a voice is chosen: on a character, in a film's cast, on a bot, and in the studio's Speech tab. The voice clone help article and the voice studio tutorial show the same flow on screen.

What reading a line costs

• In your cloned voice: about 25 credits per 1,000 characters.
• In a built-in narrator or library voice: 5 credits per started 1,000 characters.

For scale, a 60-second voiceover is roughly 140 words, under 1,000 characters, so it is one step of either price. The button shows the exact figure before you press it.

Text to speech without cloning

If the voice does not have to be yours, the Speech tab in Mint Studio runs the text to speech models directly. The prices as the studio shows them:

• Gemini 3.8 Flash-Lite TTS (the cheapest voice): about 3 credits per 1,000 characters
• Gemini 3.8 Flash TTS, 70 voices: about 5 credits per 1,000 characters
• ElevenLabs Turbo 2.5: 8 credits per 1,000 characters
• Gemini 3.1 Flash TTS: about 9 credits per 1,000 characters
• Gemini 2.5 Pro TTS: about 10 credits per 1,000 characters
• ElevenLabs Multilingual v2: 15 credits per 1,000 characters
• ElevenLabs v3, the most expressive: 18 credits per 1,000 characters

The same tab has Audio Isolation, which strips background noise from a recording, at about 15 credits per audio-minute, settled on the actual length. It is a good pass to run on a noisy sample before you clone it. The speech tutorial covers browsing voices and writing lines that read naturally.

Design a voice from a description

There is a third way to get a voice: describe it. In Voice Studio, write something like "an elderly man in his 80s, low, slightly raspy, gentle, speaks slowly with natural breaths", optionally add a line for it to say (100+ characters, ideally from your real script), audition the candidates, and save the one you like under a name. It joins your roster like a clone.

Where the voiceover goes next

• A talking photo. The take is exactly what the talking photo models take as their audio input, so "make this photo speak in my voice" is: clone once, speak, attach. See talking photo.
• A character. Give a reusable character your voice and it speaks in it in every scene it is cast in.
• A video edit. Drop the take on an audio track in Video Edit, under your clips and next to a music bed, and export.

Writing lines that sound natural

Short sentences read better than long ones. Write numbers the way they should be said. Put a full stop where you want a breath. If a word keeps coming out wrong, spell it the way it sounds. For a long narration, the AI Prompt button on a speech model writes the finished narration as flowing paragraphs at about 140 words a minute (about 110 for Arabic), so the take lands close to the length you need.

Whose voice can I clone?

Your own, or a voice you have explicit permission to use. Cloning someone without consent is against the acceptable-use policy.

Does a cloned voice work in other languages?

Yes. A clone reads 30+ languages including Arabic, Spanish, French, German, Portuguese, Italian, Turkish, Japanese, Korean, Chinese, Hindi and Russian. The accent follows the sample. Leave the language on auto-detect unless it guesses wrong.

How long should the sample be?

The clone dialog asks for 10 to 30 seconds. Clean matters more than long: one quiet, steady recording beats a minute with a fan in the background.

Can I hear a voice before I use it?

Yes. Every voice in the Voice Studio roster, built-in or your own, has a play button for an audition, so you can compare narrators before you spend anything on a line.

Can a bot narrate in my cloned voice?

Yes. A voice made in Voice Studio is pickable on a bot, on a character and in a film's cast, so a scheduled reel can be narrated in your own voice without you recording anything again.

Can I delete a voice?

Yes. Each voice in your list has a Delete action that retires it.