Voices that sound like people, not like software.

Natural Vietnamese and English text to speech, voice cloning from a ten-second sample, and an API that scales to zero. You pay per character.

Pick a voice. Press play.

Every sample below was produced by Voixa itself, from the same sentence, with no post-processing.

From text to audio in one call

Three steps, no infrastructure to run.

  1. 1

    Pick a voice

    Thirty-five built-in voices, or clone your own from a short recording.

  2. 2

    Send the text

    Up to 5 000 characters per call. Vietnamese and English, punctuation drives the pauses.

  3. 3

    Get a clip

    A WAV in your library with a signed URL, usually within seconds. Keep it, share it, delete it.

import { Voixa } from '@vovix/voixa'

const vx = new Voixa({ apiKey: process.env.VOIXA_API_KEY })

const clip = await vx.say({
  voiceId: 'en-alba',
  text: 'The lighthouse keeper switched the lamp on at dusk.',
})
console.log(clip.url) // WAV, signed URL

Zero-dependency client for Node, Bun, Deno, Workers and the browser: npm install @vovix/voixa. Read the docs.

Pay per character

No seats, no minimums, no idle servers. Accounts start with a daily allowance; ask for more when you need it.

Get started