quantum-sync-xtts-v2

XTTS-v2 Model Mirror for Quantum Sync

This is a mirror/backup of the Coqui XTTS-v2 model for use with the Quantum Sync project.

🎯 Purpose

This mirror serves as:

• Backup in case the original model becomes unavailable
• Faster access for Quantum Sync users
• Stable reference for production deployments

📋 Model Information

Original Model: coqui/XTTS-v2

Architecture: XTTS-v2 (Zero-shot multi-lingual TTS)

Model Size: ~1.87 GB

Supported Languages: 13 languages

• English (en)
• Thai (th)
• Spanish (es)
• French (fr)
• German (de)
• Italian (it)
• Portuguese (pt)
• Polish (pl)
• Turkish (tr)
• Russian (ru)
• Dutch (nl)
• Czech (cs)
• Arabic (ar)
• Chinese (zh-cn)

🚀 Usage

With Quantum Sync (Recommended)

git clone https://github.com/Useforclaude/quantum-sync-v5.git
cd quantum-sync-v5/quantum-sync-v11-production

# Configure to use this mirror
# Edit tts_engines/xtts.py, change model_name to:
# model_name = "useclaude/quantum-sync-xtts-v2"

python main_v11.py input/file.srt \
  --voice MyVoice \
  --voice-sample /path/to/voice.wav \
  --tts-engine xtts-v2 \
  --tts-language en

Direct Usage with TTS Library

from TTS.api import TTS

# Use this mirror
tts = TTS(model_name="useclaude/quantum-sync-xtts-v2")

# Generate speech
tts.tts_to_file(
    text="Hello, this is a test.",
    speaker_wav="reference_voice.wav",
    language="en",
    file_path="output.wav"
)

Voice Cloning Example

from TTS.api import TTS

# Initialize
tts = TTS(model_name="useclaude/quantum-sync-xtts-v2")

# Clone voice from reference audio (6-30 seconds)
tts.tts_to_file(
    text="The quick brown fox jumps over the lazy dog.",
    speaker_wav="my_voice_sample.wav",  # Your voice reference
    language="en",
    file_path="output_cloned.wav"
)

📊 Performance

From Quantum Sync Production Tests (2025-10-13):

Comparison:

• XTTS-v2: 15-17 min, 8/10 quality, FREE, 87% audio
• F5-TTS: 20-25 min, 7/10 quality, FREE, 55% audio
• AWS Polly: 5 min, 9/10 quality, ~$0.06, no cloning

🎛️ Advanced Parameters

# Speed control (0.5 - 2.0)
tts.tts_to_file(
    text="Hello world",
    speaker_wav="voice.wav",
    language="en",
    speed=0.8,  # Slower speech
    file_path="output.wav"
)

# Temperature control (0.1 - 1.0)
tts.tts_to_file(
    text="Hello world",
    speaker_wav="voice.wav",
    language="en",
    temperature=0.75,  # More expressive
    file_path="output.wav"
)

📦 Model Files

quantum-sync-xtts-v2/
├── model.pth              (1.87 GB - Neural network weights)
├── config.json            (Model configuration)
├── vocab.json             (Vocabulary for tokenization)
├── speakers_xtts.pth      (Speaker embeddings)
├── dvae.pth               (DVAE component)
├── mel_stats.pth          (Mel-spectrogram statistics)
├── LICENSE                (MPL 2.0)
└── README.md              (This file)

📜 License

Mozilla Public License 2.0 (MPL 2.0)

This model is licensed under the Mozilla Public License 2.0. You can:

• ✅ Use commercially (no restrictions)
• ✅ Modify the model
• ✅ Distribute the model
• ✅ Use in proprietary software

Requirements:

• Include license and copyright notice
• State changes if you modify the model
• Disclose source for modifications

Full License: LICENSE

🙏 Attribution

Original Work:

• Project: Coqui TTS
• Model: XTTS-v2
• Authors: Coqui TTS Team
• License: Mozilla Public License 2.0

This Mirror:

• Purpose: Backup for Quantum Sync project
• Maintained by: [Your Name/Organization]
• Original Source: https://huggingface.co/coqui/XTTS-v2

All credit goes to the original Coqui TTS team. This is simply a mirror for backup and convenience.

📚 Documentation

Quantum Sync Documentation:

• XTTS-v2 Quick Start Guide
• Paperspace Testing Guide

Original Documentation:

• Coqui TTS GitHub
• XTTS-v2 Paper (if available)

🔗 Links

• This Mirror: https://huggingface.co/useclaude/quantum-sync-xtts-v2
• Original Model: https://huggingface.co/coqui/XTTS-v2
• Quantum Sync Project: https://github.com/Useforclaude/quantum-sync-v5
• TTS Library: https://github.com/coqui-ai/TTS

⚠️ Disclaimer

This is an unofficial mirror maintained for backup purposes. For the latest version and official support, please refer to the original model and Coqui TTS repository.

📊 Model Card

Model Description

XTTS-v2 is a state-of-the-art zero-shot multi-lingual text-to-speech model that can clone voices from short audio samples (6-30 seconds).

Key Features:

• Zero-shot voice cloning
• Multi-lingual support (13 languages)
• High-quality natural speech
• No fine-tuning required
• Commercial use allowed

Intended Use

Primary Use Cases:

• Voice cloning for content creation
• Multi-lingual speech synthesis
• Accessibility applications
• Audiobook narration
• Video dubbing

Out-of-Scope Use:

• Impersonation without consent
• Generating misleading content
• Illegal activities

Training Data

XTTS-v2 was trained on diverse multi-lingual speech data. For details, see the original model card.

Performance

See Performance section above for detailed benchmarks from Quantum Sync project.

Ethical Considerations

Voice Cloning Ethics:

• Always obtain consent before cloning someone's voice
• Clearly label AI-generated content
• Do not use for impersonation or fraud
• Follow local regulations on synthetic media

Limitations

• May not perfectly preserve all voice characteristics
• Quality varies with reference audio quality
• Requires GPU for reasonable speed
• ~6-8 GB VRAM recommended
• Some languages may have better quality than others

🛠️ Technical Specifications

Model Type: Autoregressive Transformer-based TTS

Framework: PyTorch

Input: Text + Reference Audio (6-30 sec WAV)

Output: 24kHz WAV audio

Inference Time: ~3-5 seconds per segment (GPU)

Hardware Requirements:

• GPU: NVIDIA with CUDA support
• VRAM: 6-8 GB recommended
• RAM: 16 GB
• Disk: ~2 GB for model

Software Requirements:

• Python 3.9+
• PyTorch 2.0+
• TTS library
• CUDA 11.8+ (for GPU)

📞 Support

For this mirror:

• Issues: Quantum Sync GitHub Issues

For original model:

• Issues: Coqui TTS GitHub Issues

---

Last Updated: 2025-10-13

Mirror Version: 1.0

Model Version: XTTS-v2 (Latest as of upload date)