# XTTS-v2 - Core.Today AI API

> Coqui XTTS-v2 multilingual voice-cloning TTS. Clone a voice from a single short audio sample and speak in 16 languages — one of the most popular open-source voice cloning models.

- **Provider**: Coqui
- **Model ID**: lucataco/xtts-v2
- **Category**: Audio & TTS
- **Credits**: 26 per generation
- **Speed**: Medium
- **Quality**: Standard

## Features
- Voice cloning from one short audio sample
- 16 language support including Korean
- Cross-language cloning (speak any language in the cloned voice)
- Optional denoising for microphone recordings
- Battle-tested open-source model (7M+ runs)

## Use Cases
- Quick voice cloning experiments
- Personalized voice messages
- Multilingual dubbing prototypes
- Voice preservation projects
- Character voice generation

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **speaker**: string - Original speaker audio to clone (wav, mp3, m4a, ogg, or flv)

### Optional
- **text**: string (default: Hi there, I'm your new voice clone. Try your best to upload quality audio) - Text to synthesize in the cloned voice
- **language**: string (default: en) - Output language for the synthesized speech
  Options: en, es, fr, de, it, pt, pl, tr, ru, nl, cs, ar, zh, hu, ko, hi
- **cleanup_voice**: boolean (default: false) - Apply denoising to the speaker audio (for microphone recordings)

## Examples

### Clone and Speak Korean
Clone an English speaker and have them speak Korean
```json
{
  "model": "lucataco/xtts-v2",
  "input": {
    "speaker": "https://example.com/voice-sample.wav",
    "text": "안녕하세요, 이것은 제 목소리로 만든 한국어 음성입니다.",
    "language": "ko",
    "cleanup_voice": true
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- Use a clean 10-30 second speaker sample for the best clone quality
- Enable cleanup_voice when the sample was recorded on a phone or laptop microphone
- The cloned voice can speak any of the 16 languages regardless of the sample's language
- Billed a flat 25 credits per generation; runtime grows with text length
- Only clone voices you have permission to use

## Documentation
https://replicate.com/lucataco/xtts-v2