kgptalkie/qwen3-tts-finetuned-baraq-obama

Qwen3-TTS Fine-Tuned — Baraq Obama Voice

Fine-tuned version of Qwen3-TTS-12Hz-1.7B-Base on a custom speaker voice using SFT training.

Training Details

• Base model: Qwen/Qwen3-TTS-12Hz-1.7B-Base
• Epochs: 5 (checkpoint-epoch-4 used)
• Batch size: 4
• Learning rate: 2e-6
• Language: English

Usage

from qwen_tts import Qwen3TTSModel
import torch

model = Qwen3TTSModel.from_pretrained("kgptalkie/qwen3-tts-finetuned-baraq-obama", device_map="cuda:0", dtype=torch.bfloat16)

wavs, sr = model.generate_custom_voice(
    text="Your text here.",
    language="English",
    speaker="laxmikant",
)