Guide
Which AI voice generator to useThe best text to speech for each kind of script
The best one depends on the script: how long it is, what language it's in, how many voices you want to try and whether you want a say in how it's read. Here are ten voice models on Leap, each with a real recording and the script it read, their prices per 1,000 characters and a pick for each of those jobs.
Last updated
Read the transcript
[excited] Table nine just ordered the entire left side of the menu! [shouting] Chef! We need more limes! [panting] All of the limes!
The short answer
Which one for what
- Try Inworld TTS
- Try Inworld TTS
- Try MiniMax Speech 2.8 HD
- Try Gemini 3.8 Flash TTS
A say in how it's read
Beside the script, you write how to read it in your own words.
Side by side
Every model in this guide, spec by spec
Prices are Leap’s, in dollars. A dash means Leap’s catalog doesn’t list it.
| Model | Voices | Script per run | Price |
|---|---|---|---|
| Eleven v3ElevenLabs | 21 | Up to 5,000 characters | $0.11 per 1,000 characters |
| Eleven v4ElevenLabs | 21 | Up to 5,000 characters | $0.088 per 1,000 characters |
| MiniMax Speech 2.8 HDMiniMax | 17 | Up to 10,000 characters | $0.11 per 1,000 characters |
| Gemini 3.8 Flash TTSGoogle | 30 | Up to 4,000 characters | $0.0495 per 1,000 characters |
| MAI-Voice-2.1Microsoft | 3 | Up to 4,000 characters | $0.0242 per 1,000 characters |
| TTS-1 HDOpenAI | 9 | Up to 4,000 characters | $0.033 per 1,000 characters |
| Qwen Audio 3 TTSAlibaba | 45 | Up to 2,000 characters | $0.055 per 1,000 characters |
| Grok TTSxAI | 5 | Up to 4,000 characters | $0.0165 per 1,000 characters |
| Inworld TTSInworld | 113 | Up to 2,000 characters | $0.011 per 1,000 characters |
| Seed Speech 2ByteDance | 41 | Up to 5,000 characters | $0.033 per 1,000 characters |
The models
What each one makes
- 0 / 13 s
Read the transcript
[nervous] Hi, um, welcome to my podcast. This is episode one. [clears throat] Today we're talking about moths. [whispers] They're in the room with me right now.
A nervous first podcast, made on Leap with Eleven v3 Eleven v3
By ElevenLabs
ElevenLabs v3: expressive speech that follows audio tags like [laughs] or [whispers].
- Voices
- 21
- Script per run
- Up to 5,000 characters
- Price
- $0.11 per 1,000 characters
- 0 / 11 s
Read the transcript
Seven miles below the surface, the pressure would fold a car like paper. And yet, down here, in total darkness, something is glowing.
Documentary: seven miles down, made on Leap with Eleven v4 Eleven v4
By ElevenLabs
ElevenLabs' newest and most expressive speech model, in over 70 languages.
- Voices
- 21
- Script per run
- Up to 5,000 characters
- Price
- $0.088 per 1,000 characters
- 0 / 7 s
Read the transcript
It smells like the last train home, rain on warm stone, and a decision you will not regret in the morning.
A fragrance ad, no product named, made on Leap with MiniMax Speech 2.8 HD MiniMax Speech 2.8 HD
By MiniMax
MiniMax's studio-quality speech, with emotion and over 30 languages.
- Voices
- 17
- Script per run
- Up to 10,000 characters
- Price
- $0.11 per 1,000 characters
- 0 / 12 s
Read the transcript
Room four. The artist completed this painting in a single night, and then spent eleven years telling everyone it took a single night.
A deadpan museum audio guide, made on Leap with Gemini 3.8 Flash TTS Gemini 3.8 Flash TTS
By Google
Google's speech model: describe the delivery in style instructions.
- Voices
- 30
- Script per run
- Up to 4,000 characters
- Price
- $0.0495 per 1,000 characters
- 0 / 10 s
Read the transcript
In the high desert, the night drops forty degrees in an hour. The stones that baked all day begin to crack, and the whole valley ticks like a clock.
Documentary: the high desert at night, made on Leap with MAI-Voice-2.1 MAI-Voice-2.1
By Microsoft
Microsoft's most natural voices, for narration, across 23 languages.
- Voices
- 3
- Script per run
- Up to 4,000 characters
- Price
- $0.0242 per 1,000 characters
- 0 / 8 s
Read the transcript
Your table is ready. Please follow me, and mind the step. The chef has asked me to tell you that tonight, everything tastes better than it looks.
A maître d' with a warning, made on Leap with TTS-1 HD TTS-1 HD
By OpenAI
OpenAI's higher-quality text to speech, in nine voices.
- Voices
- 9
- Script per run
- Up to 4,000 characters
- Price
- $0.033 per 1,000 characters
- 0 / 13 s
Read the transcript
好，今天我们只做一件事: how to fold a dumpling that doesn't explode in the pot. 先拿一张皮，水不要太多，then pinch, pinch, pinch.
Dumplings, in Mandarin and English, made on Leap with Qwen Audio 3 TTS Qwen Audio 3 TTS
By Alibaba
Alibaba's speech model, strong in Chinese and English.
- Voices
- 45
- Script per run
- Up to 2,000 characters
- Price
- $0.055 per 1,000 characters
- 0 / 10 s
Read the transcript
Okay, confession: I named my sourdough starter, and now I can't throw any of it away. It's been four years. It has a birthday. We sing.
A sourdough confession, made on Leap with Grok TTS Grok TTS
By xAI
Expressive speech in five voices.
- Voices
- 5
- Script per run
- Up to 4,000 characters
- Price
- $0.0165 per 1,000 characters
- 0 / 13 s
Read the transcript
Buying or browsing? Browsing costs extra. Touching costs double. And if you breathe on the potions, we are going to have a very different conversation.
A goblin potion seller, made on Leap with Inworld TTS Inworld TTS
By Inworld
Low-cost, natural speech for characters and apps.
- Voices
- 113
- Script per run
- Up to 2,000 characters
- Price
- $0.011 per 1,000 characters
- 0 / 8 s
Read the transcript
Welcome back to Cooking at Two A.M. Tonight: instant noodles, but we are going to treat them with the respect they deserve.
Cooking at 2 a.m., made on Leap with Seed Speech 2 Seed Speech 2
By ByteDance
ByteDance's speech model, in Chinese and English.
- Voices
- 41
- Script per run
- Up to 5,000 characters
- Price
- $0.033 per 1,000 characters
Questions
Questions people ask
- What is the best text to speech?
- It depends on the recording. For the lowest price per character, Inworld TTS: $0.011 per 1,000 characters, the lowest rate here. For the most voices to choose from, Inworld TTS: 113 voices, the most here. For the longest script in one go, MiniMax Speech 2.8 HD: up to 10,000 characters in one run. For a say in how it's read, Gemini 3.8 Flash TTS: beside the script, you write how to read it in your own words. They all run in the same studio, so you can give two of them the same prompt and keep the better recording.
- Is ElevenLabs on Leap?
- Yes. Eleven v3 and Eleven v4 are in this guide, and you pay for them per character from the same balance as every other model, with no plan to pick. If you pay for ElevenLabs now, Leap vs ElevenLabs sets the two side by side.
- Can I clone my own voice?
- Not with these models: each one reads your words in its own voices, and none of them takes a recording to copy. Never make a voice that passes for a real person who hasn't agreed: Leap's Acceptable Use Policy forbids it.
- Which text to speech can I try for free?
- All of them. Sign up with a Gmail, Outlook or work address and you start with $2 of free credit, no card needed. The credit works on every model here. After that you pay for what you run, in dollars, with no subscription.
- Do I need an account with each maker?
- No. One Leap account runs all 10 of them, in the studio and through the API, from one balance in dollars. You see the price before each run, and a run that fails costs nothing.
- Can I use what I make commercially?
- Yes. You own what you make on Leap, so you can use it commercially, and we never train AI on it. The people and brands shown in a result can still have rights in it; the Terms say more.
- Is there an API?
- Yes. Every model here runs through Leap's REST API at the same price as in the studio, and agents can call them through Leap's MCP server.