Qwen3-TTS

A fine-tuned Qwen3-TTS model trained by might2901 for prompt-driven text-to-speech synthesis.

24 kHz mono WAV output, single forward call, no reference audio required.

Usage

pip install qwen-tts transformers torch soundfile

from qwen_tts import Qwen3TTSModel
import soundfile as sf

model = Qwen3TTSModel.from_pretrained("might2901/model-name")

wavs, sr = model.generate_voice_design(
    text="Hello, this is a test of the text to speech system.",
    instruct="A clear, natural voice speaking calmly.",
    language="english",
)
sf.write("output.wav", wavs[0], sr)

Prompt Guide

Layer	Examples
Gender	a man, a woman
Mood	speaking warmly, calm, natural, softly
Pace	unhurried, steady, measured
Style	conversational, professional, neutral

Example prompts:

A man speaks calmly and naturally.
A woman with a clear, conversational tone.
A professional voice, neutral and steady.

Files

model.safetensors            # model weights
speech_tokenizer/            # Qwen3 audio codec
tokenizer.json + ...         # text tokenizer
config.json                  # model config
generation_config.json       # generation settings

License

CC BY-NC-SA 4.0 — research and non-commercial use only.

Downloads last month: 29

Safetensors

Model size

2B params

Tensor type

BF16