Text to Speech voices: MiniMax, Qwen 3, OpenAI and ElevenLabs
The old split into standard and OpenAI voices is gone. The picker now has MiniMax, Qwen 3, OpenAI, ElevenLabs and your own voices, with different settings and prices.
Updated November 2, 20263 min read
The voice picker is on the Voice tab of Text to Speech. Each provider has its own tab, and two more tabs, Favorites and Recent, collect voices from all of them. The play button next to any voice plays a free sample.
How the voice groups differ
| Tab | Voices | Settings | Price | Good for |
|---|---|---|---|---|
| MiniMax | 53 voices: 45 English and 8 Russian | emotion, speed 0.5–2, pitch from −12 to +12; tags for laughter and pauses | 12 tokens per 1,000 characters | long narration, speech with emotion |
| Qwen 3 | 9 voices | none | about 20 tokens per 1,000 characters | short phrases in one of 10 languages |
| OpenAI | 10 voices | text instructions and 8 presets | 1.8 tokens per started 150 words | long texts at the lowest price, a voice with a set manner |
| ElevenLabs | Library and a large Community catalog | stability, similarity, style; tags for emotions and pauses | 12 tokens per 1,000 characters | a voice of the right gender, age, accent and language |
| My voices | voices cloned from your recordings | MiniMax voices: emotion, speed, pitch; Qwen 3 voices: none | 12 or about 11 tokens per 1,000 characters | speech in your own voice |
For comparison, a minute of speech is about 150 words or 900 characters: about 2 tokens with OpenAI, 11 with MiniMax or ElevenLabs and 18 with Qwen 3.
MiniMax voices
MiniMax has 53 ready voices with English names, for example Expressive Narrator, Calm Woman, Wise Scholar and Deep-voiced Gentleman. 45 of them are English and 8 are Russian, such as Reliable Man and Bright Queen. If the app is in Russian, the Russian voices come first in the list. Expressive Narrator is selected when you open Text to Speech.
Below the list, Voice settings appear: Emotion (Happy, Sad, Angry, Fearful, Surprised, Calm, Disgusted; tap again to turn it off), Speed from 0.5 to 2 and Pitch from −12 to +12. On the Text tab you can add tags such as (laughs) or a pause <#1#>.
Qwen 3 voices
Qwen 3 has 9 voices and no settings. They speak 10 languages: English, Russian, Chinese, German, French, Spanish, Italian, Japanese, Korean and Portuguese.
| Female | Male |
|---|---|
| Vivian, Serena, Ono Anna, Sohee | Uncle Fu, Dylan, Eric, Ryan, Aiden |
OpenAI voices
OpenAI has 10 voices. Only these voices accept instructions: a short description of how to speak, or one of 8 presets. How to write them: how to set the tone of an OpenAI voice.
| Voice | Description in the app |
|---|---|
| Alloy | Neutral, balanced, versatile |
| Ash | Male, soft, conversational |
| Ballad | Male, melodic, musical |
| Coral | Female, warm, friendly |
| Echo | Male, clear, confident |
| Fable | Male, British, storyteller |
| Nova | Female, energetic, youthful |
| Onyx | Male, deep, authoritative |
| Sage | Male, wise, calm |
| Shimmer | Female, soft, lyrical |
ElevenLabs voices
The ElevenLabs tab has two parts: Library, a set of ready voices that loads at once, and Community, a much larger catalog of voices published by other creators. Both can be filtered by language, gender, age, accent and use case. The settings are Stability, Similarity and Style, and on the Text tab there are tags such as [whispers] or [laughs].
ElevenLabs voices are also the only ones available in Voice Changer. How to search and filter them: choosing a voice.
Which voice to choose for your task
| Task | What to try first |
|---|---|
| Voice-over for a presentation or video in English | MiniMax Expressive Narrator or an OpenAI voice with the News preset |
| Text in a language other than English or Russian | ElevenLabs with the language filter |
| A long text, as cheap as possible | OpenAI |
| Speech with laughter, sighs or pauses | MiniMax or ElevenLabs with tags |
| Your own voice | My voices: see how to voice text with your own voice |
Contact support
Describe the problem and we will reply by email. Add the file name or job ID if you have one.