Qwen3TTS-0.6B Arabic Fine-Tuned (Custom Voice: Ali)

Technical Report | GitHub Repo | Base Model

This repository contains a fine-tuned version of Qwen3-TTS-12Hz-0.6B optimized for Arabic Text-to-Speech (Egyptian dialect) using a custom speaker profile (ali).

Model Details

• Base Model: Qwen3-TTS-12Hz-0.6B
• Language: Arabic (ar) / Egyptian Dialect
• Custom Speaker Key: ali
• License: Apache-2.0

Quickstart

1. Installation

pip install -U qwen-tts soundfile torch
# Optional for faster inference:
pip install -U flash-attn --no-build-isolation

import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel

# Load the model
model = Qwen3TTSModel.from_pretrained(
    "MahmoudIbrahim/Qwen3TTS0.6-ar-s-ali",
    device_map="cuda:0",
    dtype=torch.bfloat16,
    attn_implementation="flash_attention_2",
)

# Generate speech
wavs, sr = tts.generate_custom_voice(
    text="أنا كنت رايح الشغل الصبح، بس الدنيا كانت زحمة أوي واتأخرت شوية، فقولت أجيب قهوة في السكة وأكمل يومي",
    speaker="ali",
)
sf.write("output_final.wav", wavs[0], sr)