Expressive TTS API

Sensual voices, soft whispers, and intimate audio synthesis.

An AI text-to-speech API tuned for closeness. Generate breathy whispers, close-mic narration, and natural conversational turns with sub-200ms streaming latency.

Explore voice catalog View API spec

POST /v1/text-to-speech/velvet

Host: api.secretvoice.dev
xi-api-key: $SECRETVOICE_API_KEY

{
  "text": "Come closer. Let me tell you a secret. Now, just listen to my voice.",
  "model_id": "secret-v1"
}

Voice Library

Curated voices tuned for closeness.

Audition our flagship female and male vocal models below. Every voice preserves subtle breath, warmth, and micro-inflections.

Female Voices

Male Voices

Multilingual

One voice, twenty-one languages.

The same voices keep their warmth and intimacy in every language. Pass a language_code and the model speaks it natively.

POST /v1/text-to-speech/dominique

{
  "text": "Acércate. Quiero contarte un secreto.",
  "language_code": "es",
  "model_id": "secret-v1"
}

POST /v1/text-to-speech/whisper

{
  "text": "もっと近くに来て。秘密を教えてあげる。",
  "language_code": "ja",
  "model_id": "secret-v1"
}

POST /v1/text-to-speech/phoenix

{
  "text": "Approche-toi. Je vais te murmurer un secret.",
  "language_code": "fr",
  "model_id": "secret-v1"
}

Acoustic Tuning

Engineered for proximity and texture.

Whispered ASMR

Unlike standard TTS models that flatten quiet speech or introduce metallic artifacts, SecretVoice preserves soft-spoken air, delicate whispers, and close-mic vocal texture.

Sensual Vocal Styles

Natural breath control, unhurried pacing, and chest-voice resonance that bring emotional weight and physical closeness to every phrase.

Expressive AI Narration

Context-aware prosody across long passages, built for romantic fiction, interactive audio stories, and immersive character performance.

Real-time Voice Agents

Built for live, low-latency conversation.

Power AI companions and interactive voice agents with streaming audio that responds immediately and never breaks character.

WebSocket & HTTP Chunked Streaming

Stream audio frames directly from LLM token output with sub-200ms time-to-first-audio for instant conversational turn-taking.

Stable Persona Across Turns

Zero acoustic drift over multi-turn sessions. Voice timbre, intimacy level, and pacing remain locked from the first word to the thousandth.

Drop-In REST & WebSocket API

Simple JSON payload with PCM, MP3, or Opus output formats, ready to pipe straight into WebRTC, LiveKit, or browser audio contexts.

WS /v1/text-to-speech/velvet/stream-input

wss://ws.secretvoice.dev/v1/text-to-speech/velvet/stream-input
xi-api-key: $SECRETVOICE_API_KEY

# send
{ "text": " " }
{ "text": "There you are. I was starting to miss you." }
{ "flush": true }
{ "text": "" }

# recv
{ "audio": "<base64>", "isFinal": false }
{ "isFinal": true }