Text to speech with human-like voices
Natural, expressive speech from text in 25+ languages.
Carina
Fine-grained control
Control emotion, pacing, and emphasis with simple speech tags.
- Style tags to control delivery.
- REST for long-form audio, WebSocket for real-time streaming.
25+ languages
Every voice speaks 25+ languages with natural intonation.
- Voices keep their character across languages.
- Switch languages seamlessly.
Close your eyes and listen for a moment. Every voice tells a story, and every story deserves to be heard. Wherever you are in the world, I'm glad you're here.
Human-like, expressive speech
Built on the same stack that powers HypeV Voice agents and customer support.
- Ranked closest to human on the Vapi Humanness Index.
- Podcasts, voiceovers, audiobooks, characters, and support agents.
Vapi Humanness IndexAs of July 10, 2026
Simple, transparent pricing
Usage-based billing with no hidden fees.
- A fraction of competitors' prices.
- Pay for what you use.
Price per 1M characters
Clone your voice
Clone a voice from two minutes of speech and use it instantly.
- 30 free clones to start.
- Works with speech tags and all 25+ languages.
Speed
Speed up or slow down delivery.
Output formats
MP3, WAV, and raw PCM, plus μ-law and A-law for telephony.
Text normalization
Numbers, dates, and symbols read the way they're spoken.
Word timestamps
Character-level start and end times for captions, karaoke highlights, and lip-sync.
Streaming optimization
Three latency levels trade time-to-first-audio against quality.
Pronunciation overrides
Spell out how names, brands, and jargon should sound.
Sample rate
From 8 kHz telephony to 48 kHz studio audio.
Bit rate
MP3 output from 32 up to 192 kbps.