AiVoxta is a next-generation neural text-to-speech studio powered by Fish Audio S2.1, supporting expressive emotion tags, multilingual voice generation in Hindi, English, and Japanese, and zero-shot voice cloning.

How do Emotion Tags work in AiVoxta?

You can insert natural text tags like [whispering], [excited], [dramatic], or [sighs] directly in your script without complex XML. The neural model naturally adjusts pitch, cadence, and breath in real time.

Is my data private with AiVoxta?

Yes. AiVoxta uses a client-first BYOK (Bring Your Own Key) architecture where your API credentials and audio generation stream directly without being permanently stored on intermediary servers.

Home/Docs/Developer Resources

AiVoxta API v1.0OpenAPI 3.1.0 ReadyMCP Protocol 2024-11-05

AiVoxta Developer Resources & API Documentation

Integrate studio-grade neural text-to-speech powered by Fish Audio S2.1 into applications, autonomous agents, and workflows. Exposed via predictable REST endpoints, an OpenAPI 3.1.0 specification, and a live Streamable HTTP Model Context Protocol (MCP) server.

OpenAPI 3.1.0

openapi.json

Machine-readable JSON schema for API generators and Swagger/Postman.

MCP Live Handshake

/.well-known/mcp

Model Context Protocol manifest and JSON-RPC streamable tools for agents.

Agent Guidance

llms.txt

Standardized llmstxt.org index specifying when and how agents should call AiVoxta.

Extended Context

llms-full.txt

Full API parameters, emotion tags dictionary, and tool schema reference.

Authentication & Privacy Architecture (BYOK)

AiVoxta implements a client-first Bring-Your-Own-Key model. Authenticate API requests to protected endpoints by supplying your personal Fish Audio API key via either HTTP request header:

Header 1 (Recommended)

x-fish-api-key: your_api_key_here

Header 2 (Standard Bearer)

Authorization: Bearer your_api_key_here

Don't have an API key yet? Follow our Free APIs Guide to acquire complimentary Fish Audio generation credits.

Core REST API Endpoints

POST/api/tts

Returns: audio/mpeg or audio/wav

Synthesize high-fidelity audio from text. Supports inline performance tags like [whispering], [excited], [dramatic] without SSML.

# Example cURL Synthesis

curl -X POST https://aivoxta.vercel.app/api/tts \
  -H "Content-Type: application/json" \
  -H "x-fish-api-key: $FISH_API_KEY" \
  -d '{
    "text": "Welcome to AiVoxta. [excited] Let us build amazing voice agents!",
    "voice_id": "hindi-female",
    "format": "mp3",
    "temperature": 0.7
  }' --output voice.mp3

GET/api/voices

Returns: application/json

Fetch directory of curated and dynamic neural voices across Hindi, English, and Japanese.

curl https://aivoxta.vercel.app/api/voices

POST/api/enhance

Returns: application/json

Automatically annotate plain dialogue with expressive emotion tags using LLMs or the built-in rule-based tagger.

GET/api/wallet

Requires: x-fish-api-key

Query current Fish Audio credit balance and active subscription package.

Model Context Protocol (MCP) Server Setup

AiVoxta exposes an official Model Context Protocol server via Streamable HTTP transport. Connect Claude Desktop, Cursor, ChatGPT, or custom agent runtimes by adding the endpoint to your configuration:

# claude_desktop_config.json or agent mcp settings

{
  "mcpServers": {
    "aivoxta": {
      "url": "https://aivoxta.vercel.app/.well-known/mcp"
    }
  }
}

synthesize_speechGenerates audio binary or preparation metadata from script text and voice ID.

list_voicesReturns all curated and custom voices filtered by language or category.

enhance_promptInjects emotion tags like [whispering], [dramatic] into script text.

check_walletQueries Fish Audio API credit balance and tier validity.

Emotion Performance Tags Dictionary

Insert any of the following natural tags directly within your script text to shape pitch, cadence, and breath:

[whispering]

[excited]

[dramatic]

[sighs]

[laughter]

[gasp]

[pause]

[clears throat]

[calm]

[curious]

[angry]

[hesitant]