My Voices: what a custom voice is and where it works
A custom voice is a copy of your voice made from a 10 s to 5 min recording. It speaks in Text to Speech and Translate audio on every device you sign in to.
Updated November 2, 20265 min read
All your voices are listed in Settings → My Voices. Each one is made by one of two providers, MiniMax or Qwen 3. You pick the provider before you save the recording, and the badge next to the voice name shows which one it is.
Where you can use your voice
A custom voice works in Text to Speech, in Translate audio and in AI agents connected over MCP. Voice Changer, AI Calls and Translate Video don't use saved voices.
| Tool | Your voice | How it works there |
|---|---|---|
| Text to Speech | yes | The My voices tab in the voice picker. See how to voice text with your own voice. |
| Translate audio | yes | Settings → Voiceover → MiniMax or Qwen 3 → My MiniMax voice or My Qwen voice. The translation is read in your voice. |
| AI agents (MCP) | yes | Claude, ChatGPT and other agents see the same list of voices. See connecting VocaLingo to AI agents. |
| Voice Changer | no | Only voices from the ElevenLabs library and community. |
| AI Calls | no | The translation in a call is spoken by a standard voice. You can't pick your own. |
| Translate Video | no | The voice is cloned from the speaker in the video itself. |
MiniMax or Qwen 3: which provider to pick
MiniMax costs more to create and speaks more languages. Qwen 3 is almost free to create and speaks 10 languages. Both need the same recording: from 10 seconds to 5 minutes, up to 20 MB, 60 seconds recommended.
| MiniMax | Qwen 3 | |
|---|---|---|
| Creating the voice | 180 tokens, once | about 1 token |
| Voicing text | 12 tokens per 1,000 characters | about 11 tokens per 1,000 characters |
| Languages | 38 in Translate audio | 10: English, Russian, Chinese, German, French, Spanish, Italian, Japanese, Korean, Portuguese |
| Settings in Text to Speech | emotion, speed, pitch | none |
MiniMax gives better results when the recording is in the same language you plan to voice. If you are not sure, start with Qwen 3: a second voice from the same recording costs about 1 token.
How to get a custom voice
The main way is to record or upload a sample in My Voices: step by step in how to clone your voice. A voice also appears in the list in these cases:
You recorded your voice in the welcome tour. That voice is saved for free as a Qwen 3 voice.
After a translation in Translate audio that cloned the voice from the file, you tapped Save voice under What next?. Saving a MiniMax clone this way costs nothing extra, because the translation already paid for it.
In the chat you attached an audio file and tapped the Clone voice chip. The chat always makes a Qwen 3 voice.
An AI agent saved a voice for you with the save_voice tool. It also makes a Qwen 3 voice.
How much it costs to clone a voice and voice a minute
A MiniMax voice costs 180 tokens and a Qwen 3 voice about 1 token. A minute of speech is about 150 words, or roughly 900 characters of English text, so voicing it in your voice costs about 10–11 tokens with either provider.
The 180 tokens for MiniMax are charged when cloning starts. If cloning fails, they come back to your balance.
Your balance needs at least 180 tokens for a MiniMax voice. With less, the voice ends with the Failed badge.
Premium isn't required. Prices are the same with and without it.
Prices for all tools: token pricing for each tool.
Will my voice move to another phone or computer
Yes. Voices are stored in your VocaLingo account. Sign in to the same account on another phone or on web.vocalingo.app, and the same voices are in My Voices. Sign in with Apple works only on iOS, so on Android and the web use the email or Google sign-in of that account: see how to create an account and sign in.
If you created a voice as a guest and then linked the guest to an email, Google or Apple, the voice stays.
If you created a voice as a guest and then signed in to another account that already existed, only the token balance moves to that account. The voice stays with the guest.
Favorites and Recent in the voice picker are kept on the device and are cleared when you log out. Your voices themselves are not affected.
Whose voice you may clone
Clone only your own voice or the voice of a person who has agreed to it. The Privacy Policy allows only voices you own or hold the rights and consent for, and forbids using a synthetic voice to impersonate or deceive anyone. VocaLingo may delete voices and results made with them when the rights are missing. How the recording is processed: how VocaLingo processes your data.
Frequently asked questions
The app says a MiniMax voice is deleted if unused for 7 days. Is that true?
Not for new voices. A MiniMax voice is activated right when it is created, so it stays until you delete it. The hint on the New voice screen is outdated.
How many voices can I save?
The app doesn't limit the number of voices. Each new voice is paid separately.
Can I edit a voice after it's created?
You can only rename it. To improve the sound, create a new voice from a better recording and delete the old one.
Contact support
Describe the problem and we will reply by email. Add the file name or job ID if you have one.