Private preview · POST /v1/text-to-speech

Build voice AI that
talks back in real time.

One API for lifelike speech, accurate transcription, natural turn-taking and phone-ready voice agents, running on our own infrastructure.

Test the voice

Sample line117 / 280

Pick a sample line or type your own, choose a voice, and press Play.

Performance

Fast enough for a real conversation.

Example≈200 mstime to first audio, text to speech
Example≈300 msend-of-turn detection
Example<1 svoice agent reply, caller stops to agent speaks
Today8voices across US and UK English

Example figures until we publish measured benchmarks.

Developers

Your first voice in one request.

Plain HTTPS and JSON. Send text, get audio. Create an agent, open a session, stream audio both ways.

  • One API key for speech, transcription and agents
  • Streaming over WebSocket for live audio
  • Usage and cost reported per request
  • AI disclosure built into every agent
Preview API
curl -X POST https://api.voxvane.com/v1/text-to-speech/aria \
  -H "Authorization: Bearer $VOXVANE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello from VoxVane.", "output_format": "mp3"}' \
  --output hello.mp3

Products

Everything a voice needs, as building blocks.

Use one piece or the whole stack. Every part speaks the same API. Products marked Preview are in private preview; Coming soon means not available yet.

Phone agent demo

A voice agent on a real phone line.

An example use case: an after-hours receptionist for a fictional plumbing company. Press play and watch the transcript and structured output fill in live.

0:000:53
Northgate PlumbingReady · 0:53 call

    Voices

    Eight voices, each with its own colour.

    Press play to hear each voice introduce itself.

    • Aria

      US English

    • Brooke

      US English

    • Nora

      US English

    • Sienna

      US English

    • Miles

      US English

    • Felix

      US English

    • Elise

      UK English

    • Graham

      UK English

    Browse all voices

    Responsible by default

    Built so people know they are talking to an AI.

    Disclosure in every agent

    Agents require a disclosure line and answer truthfully when asked whether they are an AI.

    Consent for custom voices

    A custom voice needs the speaker's recorded consent and is watermarked on every use.

    Open-source, credited

    We build on open models with commercial licences, run them ourselves, and credit every project.

    Start building with VoxVane.

    The API is in private preview. Read the quickstart, then ask for a key.