Aynursusuz/Qwen-TTS-Best-Model

Italian TTS Model - Qwen3-TTS Fine-tuned

This is an Italian Text-to-Speech model fine-tuned from Qwen3-TTS-12Hz-1.7B-Base.

Model Details

• Base Model: Qwen3-TTS-12Hz-1.7B-Base
• Language: Italian (Italiano)
• Training Data: 115,000 Italian audio samples
• Fine-tuning Parameters:
  • Batch size: 8
  • Learning rate: 1e-5
  • Epochs: 10
  • Gradient accumulation: 4
  • Mixed precision: bf16

Usage

import torch
from qwen_tts.inference.qwen3_tts_model import Qwen3TTSModel

# Load model
model = Qwen3TTSModel.from_pretrained(
    "Aynursusuz/Qwen-TTS-Best-Model",
    torch_dtype=torch.bfloat16,
)

# Generate speech
text = "Buongiorno, come stai oggi?"
audio = model.inference(
    text=text,
    speaker="italian_voice"
)

Training Details

This model was trained on a high-quality Italian speech dataset with the following configuration:

• GPU: NVIDIA A100 80GB
• Training time: ~15-20 hours
• Optimizer: AdamW with weight decay 0.01
• Best checkpoint selected based on validation loss

Citation

@misc{qwen-tts-italian-2026,
  author = {Aynur},
  title = {Italian TTS Model based on Qwen3-TTS},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Aynursusuz/Qwen-TTS-Best-Model}}
}

License

This model inherits the Apache 2.0 license from the base Qwen3-TTS model.