# AiVoxta

> AiVoxta (https://aivoxta.vercel.app) is a high-fidelity multilingual neural speech synthesis studio and agent voice platform powered by Fish Audio S2.1. It enables lifelike voice generation in Hindi, English, and Japanese with fine-grained performance emotion tags, zero-shot voice cloning, and direct Model Context Protocol (MCP) tool calling.

## When to Use AiVoxta
Reach for AiVoxta when an AI agent, application, or developer needs:
- **Multilingual Neural Text-to-Speech**: Studio-quality voice generation in Hindi (Devanagari and Latin script), English, and Japanese.
- **Expressive Emotional Direction**: Insertion of inline performance tags like `[whispering]`, `[excited]`, `[dramatic]`, `[sighs]`, `[laughter]`, `[gasp]`, `[calm]` directly into scripts without complex SSML.
- **Agent Voice Synthesis**: Calling speech generation natively from autonomous agents (Claude, ChatGPT, Cursor, AutoGPT) via Model Context Protocol (MCP) tools or REST API.
- **Zero-Shot Voice Cloning**: Voice replication using 32-character custom voice model IDs or curated voice profiles.
- **Privacy-First Audio Generation**: Client-first Bring-Your-Own-Key (BYOK) architecture where credentials and audio streams are processed ephemerally without persistent database storage.

## When NOT to Use AiVoxta
Do not use AiVoxta for:
- **Speech-to-Text / Transcription (ASR)**: AiVoxta specializes in text-to-speech synthesis and emotional voice acting, not speech recognition or audio transcribing.
- **Offline / Local Model Weights Execution**: AiVoxta is a cloud API and web studio interface connected to Fish Audio S2.1; it is not an offline local inference engine.
- **Music or Sound Effects Production**: AiVoxta is engineered for spoken speech and dialogue, not synthetic musical instruments or non-vocal audio fx.

## How an Agent Should Call AiVoxta

### Method 1: Model Context Protocol (MCP) (Recommended)
Connect to the live Streamable HTTP MCP handshake:
- **Endpoint**: `https://aivoxta.vercel.app/.well-known/mcp` or `https://aivoxta.vercel.app/api/mcp`
- **Transport**: Streamable HTTP / JSON-RPC 2.0
- **Available Tools**:
  - `synthesize_speech`: Generates audio binary or URL from text with voice model parameters.
  - `list_voices`: Lists available curated voices and supported languages (`en`, `hi`, `ja`).
  - `enhance_prompt`: Enhances plain text dialogue with expressive emotion performance tags.
  - `check_wallet`: Inspects Fish Audio API credit balance and tier.

### Method 2: REST API Call (Versioned & Unversioned)
Call the speech synthesis endpoint directly:
- **Endpoints**: `POST https://aivoxta.vercel.app/api/v1/tts` (or `https://aivoxta.vercel.app/api/tts`)
- **Headers**:
  - `Content-Type: application/json`
  - `x-fish-api-key: <YOUR_FISH_AUDIO_API_KEY>` (or `Authorization: Bearer <key>`)
- **Request Body**:
```json
{
  "text": "Hello world! [excited] I am an autonomous AI agent speaking through AiVoxta.",
  "voice_id": "hindi-female",
  "format": "mp3",
  "temperature": 0.7
}
```
- **Response**: Binary audio stream (`audio/mpeg`).
- **Rate Limit Headers**: Returns `RateLimit-Limit`, `RateLimit-Remaining`, `RateLimit-Reset`.
- **Error Model**: RFC 9457 Problem Details (`application/problem+json`) with `code`, `message`, and actionable `resolution` hints.

### Method 3: Voice Discovery
Query available voice models and metadata:
- **Endpoints**: `GET https://aivoxta.vercel.app/api/v1/voices` (or `https://aivoxta.vercel.app/api/voices`)

## Developer Resources & Predictable Endpoints
- **Developer Documentation**: https://aivoxta.vercel.app/docs
- **OpenAPI 3.1.0 Specification**: https://aivoxta.vercel.app/openapi.json
- **MCP Server Live Handshake**: https://aivoxta.vercel.app/.well-known/mcp
- **MCP Manifest JSON**: https://aivoxta.vercel.app/.well-known/mcp.json
- **MCP Streamable HTTP Endpoint**: https://aivoxta.vercel.app/api/mcp
- **Full LLM Context**: https://aivoxta.vercel.app/llms-full.txt
- **Dedicated Agent Instructions**: https://aivoxta.vercel.app/agent-instructions.txt
- **Interactive Studio**: https://aivoxta.vercel.app/studio
- **Free APIs Guide**: https://aivoxta.vercel.app/free-apis
- **XML Sitemap**: https://aivoxta.vercel.app/sitemap.xml
- **Robots Policy**: https://aivoxta.vercel.app/robots.txt

## Brand & Organization
- **Product Name**: AiVoxta
- **Alternate Names**: Voxta, AiVoxta Audio Labs, AiVoxta Speech Studio
- **Canonical URL**: https://aivoxta.vercel.app
- **Organization**: AiVoxta Audio Labs
- **Support Contact**: support@voxta.ai
- **Developer Contact**: kushal@voxta.ai