Skip to content

GitLab

Commit e3cdb305 authored by weishb's avatar weishb
Browse files

Add Qwen3-TTS

parents
Pipeline #3390 failed with stages
in 0 seconds
Hide whitespace changes
Inline Side-by-side
Showing with 2108 additions and 0 deletions
.gitattributes 0 → 100644
View file @ e3cdb305
# 统一所有文本文件使用 LF 换行符（Linux 风格）
* text=auto eol=lf
# 显式标记二进制文件（避免误判）
*.png binary
*.jpg binary
*.wav binary
*.whl binary
*.zip binary
\ No newline at end of file
LICENSE.txt 0 → 100644
View file @ e3cdb305
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright 2026 Alibaba Cloud
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
README.md 0 → 100644
View file @ e3cdb305
# Qwen3-TTS
## 论文
[Qwen3-TTS Technical Report](https://arxiv.org/abs/2601.15621)
## 模型简介
由通义千问（Qwen）开发的一系列强大的语音生成能力，全面支持声音克隆、音色设计、超高质量拟人化语音合成以及基于自然语言的语音控制，为开发者和用户提供了目前最丰富的语音生成功能集。
<div align=center>
<img src="./doc/qwen3-tts.png"/>
</div>
Qwen3-TTS 覆盖10种主要语言（中文、英文、日文、韩文、德文、法文、俄文、葡萄牙文、西班牙文和意大利文），并提供多种方言音色配置，以满足全球化的应用需求。此外，该模型具备强大的上下文理解能力，可根据指令和文本语义自适应调节语调、语速和情感表达，并对含噪声的输入文本展现出显著增强的鲁棒性。
主要特性如下：
强大的语音表征能力：基于自研的 Qwen3-TTS-Tokenizer-12Hz，实现对语音信号的高效声学压缩与高维语义建模，完整保留副语言信息（如语气、情绪）及声学环境特征，并通过轻量级非 DiT 架构实现高速、高保真的语音重建。
通用端到端架构：采用离散多码本语言模型（LM）架构，实现全信息端到端语音建模，彻底规避了传统“语言模型 + DiT”方案中存在的信息瓶颈与级联误差问题，显著提升模型的通用性、生成效率和性能上限。
极致低延迟流式生成：基于创新的双轨混合流式生成架构，单个模型同时支持流式与非流式生成模式。在用户仅输入单个字符后即可立即输出首个音频包，端到端合成延迟低至 97 毫秒，充分满足实时交互场景的严苛要求。
智能文本理解与语音控制：支持由自然语言指令驱动的语音生成，可灵活调控音色、情感、韵律等多维度声学属性。通过深度融合文本语义理解能力，模型能自适应调整语调、节奏与情感表达，实现“所想即所听”的拟人化语音输出。
## 环境依赖
- 列举基础环境需求，根据实际情况填写
| 软件 | 版本 |
| :------: | :------: |
| DTK | 25.04.2 |
| python | 3.10.12 |
| transformers | 4.57.3 |
| vllm | 0.9.2+das.opt2.dtk25042 |
| torchaudio | 2.5.1+das.opt1.dtk25042.20251127.g10a9ffcd |
| transformer_engine | 2.5.0+das.opt1.dtk25042 |
推荐使用镜像:harbor.sourcefind.cn:5443/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk26.04-0130-py3.10-20260202
- 挂载地址`-v`，`{docker_name}`和 `{docker_image_name}`根据实际模型情况修改
```bash
docker run -it \
--shm-size 60g \
--network=host \
--name qwen3-tts \
--privileged \
--device=/dev/kfd \
--device=/dev/dri \
--device=/dev/mkfd \
--group-add video \
--cap-add=SYS_PTRACE \
--security-opt seccomp=unconfined \
-u root \
-v /opt/hyhal/:/opt/hyhal/:ro \
-v /path/your_code_data/:/path/your_code_data/ \
harbor.sourcefind.cn:5443/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.2-1226-das1.7-py3.10-20251226 bash
```
更多镜像可前往[光源](https://sourcefind.cn/#/service-list)下载使用。
关于本项目DCU显卡所需的特殊深度学习库可从[光合](https://developer.sourcefind.cn/tool/)开发者社区下载安装
其它包参照requirements.txt安装：
```
pip install -r requirements.txt
```
镜像内其他环境配置
```
1.重新安装torchaudio
pip uninstall torchaudio
pip install torchaudio-2.5.1+das.opt1.dtk25042.20251127.g10a9ffcd-cp310-cp310-manylinux_2_28_x86_64.whl
2.解压vllm.zip到/usr/local/lib/python3.10/dist-packages直接覆盖需要修改的文件
unzip -o vllm.zip -d /usr/local/lib/python3.10/dist-packages
```
## 数据集
暂无
## 训练
暂无
## 推理
### transformers
#### 单机推理
```
VoiceDesign推理
python test_model_12hz_voice_design.py
CustomVoice
python test_model_12hz_custom_voice.py
Voice Clone
python test_model_12hz_base.py
```
### vllm
#### 单机推理（以VoiceDesign为例子，CustomVoice和Voice Clone需要切换模型）
启动服务
```bash
VLLM_USE_V1=0 python -m vllm.entrypoints.openai.api_server --model Qwen3-TTS/Qwen3-TTS-12Hz-1.7B-VoiceDesign --served-model-name qwen3-tts --host 0.0.0.0 --port 8000 --trust-remote-code --dtype bfloat16 --disable-async-output-proc
```
调用服务：
```
VoiceDesign
curl -sS http://127.0.0.1:8000/v1/audio/speech \
-H "Content-Type: application/json" \
-o output.wav \
-d '{
"model":"qwen3-tts",
"text":"哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"task_type":"VoiceDesign",
"language":"Auto",
"instruct":"体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
"generation_params":{
"max_new_tokens":4096,
"do_sample":true,
"top_k":50,
"top_p":1.0,
"temperature":0.9
},
"response_format":"wav"
}'
CustomVoice
curl -sS http://127.0.0.1:8000/v1/audio/speech \
-H "Content-Type: application/json" \
-o output_customvoice.wav \
-d '{
"model":"qwen3-tts",
"text":"哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"task_type":"CustomVoice",
"speaker":"YourSpeakerName",
"language":"Auto",
"instruct":"",
"generation_params":{
"max_new_tokens":4096,
"do_sample":true,
"top_k":50,
"top_p":1.0,
"temperature":0.9
},
"response_format":"wav"
}'
Voice Clone
curl -sS http://127.0.0.1:8000/v1/audio/speech \
-H "Content-Type: application/json" \
-o output_clone_icl.wav \
-d '{
"model":"qwen3-tts",
"text":"今天的风很温柔，我们一起出去走走吧。",
"task_type":"Base",
"language":"Auto",
"ref_audio":"/path/to/ref.wav",
"ref_text":"参考音频对应的文本内容",
"x_vector_only_mode":false,
"generation_params":{
"max_new_tokens":4096,
"do_sample":true,
"top_k":50,
"top_p":1.0,
"temperature":0.9
},
"response_format":"wav"
}'
```
## 效果展示
示例输出音频：output_audio\output.wav
### 精度
`DCU与GPU精度一致，推理框架：vllm`
## 预训练权重
| 模型名称 | 权重大小 | DCU型号 | 最低卡数需求 |下载地址|
|:-----:|:----------:|:----------:|:---------------------:|:----------:|
| Qwen3-TTS-12Hz-1.7B-VoiceDesign | 1.7B | K100AI | 1 | [Modelscope] https://www.modelscope.cn/models/Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign|
| Qwen3-TTS-12Hz-1.7B-CustomVoice | 1.7B | K100AI | 1 | [Modelscope] https://www.modelscope.cn/models/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice|
| Qwen3-TTS-12Hz-1.7B-Base | 1.7B | K100AI | 1 | [Modelscope] https://www.modelscope.cn/models/Qwen/Qwen3-TTS-12Hz-1.7B-Base|
## 源码仓库及问题反馈
- https://developer.sourcefind.cn/codes/weishb/qwen3-tts_pytorch
## 参考资料
- https://github.com/QwenLM/Qwen3-TTS
\ No newline at end of file
README_origin.md 0 → 100644
View file @ e3cdb305
---
license: apache-2.0
---
# Qwen3-TTS
## Overview
### Introduction
<p align="center">
<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/qwen3_tts_introduction.png" width="90%"/>
<p>
Qwen3-TTS covers 10 major languages (Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian) as well as multiple dialectal voice profiles to meet global application needs. In addition, the models feature strong contextual understanding, enabling adaptive control of tone, speaking rate, and emotional expression based on instructions and text semantics, and they show markedly improved robustness to noisy input text. Key features:
* **Powerful Speech Representation**: Powered by the self-developed Qwen3-TTS-Tokenizer-12Hz, it achieves efficient acoustic compression and high-dimensional semantic modeling of speech signals. It fully preserves paralinguistic information and acoustic environmental features, enabling high-speed, high-fidelity speech reconstruction through a lightweight non-DiT architecture.
* **Universal End-to-End Architecture**: Utilizing a discrete multi-codebook LM architecture, it realizes full-information end-to-end speech modeling. This completely bypasses the information bottlenecks and cascading errors inherent in traditional LM+DiT schemes, significantly enhancing the model’s versatility, generation efficiency, and performance ceiling.
* **Extreme Low-Latency Streaming Generation**: Based on the innovative Dual-Track hybrid streaming generation architecture, a single model supports both streaming and non-streaming generation. It can output the first audio packet immediately after a single character is input, with end-to-end synthesis latency as low as 97ms, meeting the rigorous demands of real-time interactive scenarios.
* **Intelligent Text Understanding and Voice Control**: Supports speech generation driven by natural language instructions, allowing for flexible control over multi-dimensional acoustic attributes such as timbre, emotion, and prosody. By deeply integrating text semantic understanding, the model adaptively adjusts tone, rhythm, and emotional expression, achieving lifelike “what you imagine is what you hear” output.
### Model Architecture
<p align="center">
<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/overview.png" width="80%"/>
<p>
### Released Models Description and Download
Below is an introduction and download information for the Qwen3-TTS models that have already been released. Other models mentioned in the technical report will be released in the near future. Please select and download the model that fits your needs.
| Tokenizer Name | Description |
|---------------------------------|-------------|
| Qwen3-TTS-Tokenizer-12Hz | The Qwen3-TTS-Tokenizer-12Hz model which can encode the input speech into codes and decode them back into speech. |
| Model | Features | Language Support | Streaming | Instruction Control |
|---|---|---|---|---|
| Qwen3-TTS-12Hz-1.7B-VoiceDesign | Performs voice design based on user-provided descriptions. | Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian | ✅ | ✅ |
| Qwen3-TTS-12Hz-1.7B-CustomVoice | Provides style control over target timbres via user instructions; supports 9 premium timbres covering various combinations of gender, age, language, and dialect. | Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian | ✅ | ✅ |
| Qwen3-TTS-12Hz-1.7B-Base | Base model capable of 3-second rapid voice clone from user audio input; can be used for fine-tuning (FT) other models. | Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian | ✅ | |
| Qwen3-TTS-12Hz-0.6B-CustomVoice | Supports 9 premium timbres covering various combinations of gender, age, language, and dialect. | Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian | ✅ | |
| Qwen3-TTS-12Hz-0.6B-Base | Base model capable of 3-second rapid voice clone from user audio input; can be used for fine-tuning (FT) other models. | Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian | ✅ | |
During model loading in the qwen-tts package or vLLM, model weights will be automatically downloaded based on the model name. However, if your runtime environment is not conducive to downloading weights during execution, you can refer to the following commands to manually download the model weights to a local directory:
```bash
# Download through ModelScope (recommended for users in Mainland China)
pip install -U modelscope
modelscope download --model Qwen/Qwen3-TTS-Tokenizer-12Hz --local_dir ./Qwen3-TTS-Tokenizer-12Hz
modelscope download --model Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --local_dir ./Qwen3-TTS-12Hz-1.7B-CustomVoice
modelscope download --model Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign --local_dir ./Qwen3-TTS-12Hz-1.7B-VoiceDesign
modelscope download --model Qwen/Qwen3-TTS-12Hz-1.7B-Base --local_dir ./Qwen3-TTS-12Hz-1.7B-Base
modelscope download --model Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice --local_dir ./Qwen3-TTS-12Hz-0.6B-CustomVoice
modelscope download --model Qwen/Qwen3-TTS-12Hz-0.6B-Base --local_dir ./Qwen3-TTS-12Hz-0.6B-Base
# Download through Hugging Face
pip install -U "huggingface_hub[cli]"
huggingface-cli download Qwen/Qwen3-TTS-Tokenizer-12Hz --local-dir ./Qwen3-TTS-Tokenizer-12Hz
huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --local-dir ./Qwen3-TTS-12Hz-1.7B-CustomVoice
huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign --local-dir ./Qwen3-TTS-12Hz-1.7B-VoiceDesign
huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-Base --local-dir ./Qwen3-TTS-12Hz-1.7B-Base
huggingface-cli download Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice --local-dir ./Qwen3-TTS-12Hz-0.6B-CustomVoice
huggingface-cli download Qwen/Qwen3-TTS-12Hz-0.6B-Base --local-dir ./Qwen3-TTS-12Hz-0.6B-Base
```
## Quickstart
### Environment Setup
The easiest way to quickly use Qwen3-TTS is to install the `qwen-tts` Python package from PyPI. This will pull in the required runtime dependencies and allow you to load any released Qwen3-TTS model. We recommend using a **fresh, isolated environment** to avoid dependency conflicts with existing packages. You can create a clean Python 3.12 environment like this:
```bash
conda create -n qwen3-tts python=3.12 -y
conda activate qwen3-tts
```
then run:
```bash
pip install -U qwen-tts
```
If you want to develop or modify the code locally, install from source in editable mode.
```bash
git clone https://github.com/QwenLM/Qwen3-TTS.git
cd Qwen3-TTS
pip install -e .
```
Additionally, we recommend using FlashAttention 2 to reduce GPU memory usage.
```bash
pip install -U flash-attn --no-build-isolation
```
If your machine has less than 96GB of RAM and lots of CPU cores, run:
```bash
MAX_JOBS=4 pip install -U flash-attn --no-build-isolation
```
Also, you should have hardware that is compatible with FlashAttention 2. Read more about it in the official documentation of the [FlashAttention repository](https://github.com/Dao-AILab/flash-attention). FlashAttention 2 can only be used when a model is loaded in `torch.float16` or `torch.bfloat16`.
### Python Package Usage
After installation, you can import `Qwen3TTSModel` to run custom voice TTS, voice design, and voice clone. The model weights can be specified either as a Hugging Face model id (recommended) or as a local directory path you downloaded. For all the `generate_*` functions below, besides the parameters shown and explicitly documented, you can also pass generation kwargs supported by Hugging Face Transformers `model.generate`, e.g., `max_new_tokens`, `top_p`, etc.
#### Custom Voice Generate
For custom voice models (`Qwen3-TTS-12Hz-1.7B/0.6B-CustomVoice`), you just need to call `generate_custom_voice`, passing a single string or a batch list, along with `language`, `speaker`, and optional `instruct`. You can also call `model.get_supported_speakers()` and `model.get_supported_languages()` to see which speakers and languages the current model supports.
```python
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
model = Qwen3TTSModel.from_pretrained(
"Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
device_map="cuda:0",
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# single inference
wavs, sr = model.generate_custom_voice(
text="其实我真的有发现，我是一个特别善于观察别人情绪的人。",
language="Chinese", # Pass `Auto` (or omit) for auto language adaptive; if the target language is known, set it explicitly.
speaker="Vivian",
instruct="用特别愤怒的语气说", # Omit if not needed.
)
sf.write("output_custom_voice.wav", wavs[0], sr)
# batch inference
wavs, sr = model.generate_custom_voice(
text=[
"其实我真的有发现，我是一个特别善于观察别人情绪的人。",
"She said she would be here by noon."
],
language=["Chinese", "English"],
speaker=["Vivian", "Ryan"],
instruct=["", "Very happy."]
)
sf.write("output_custom_voice_1.wav", wavs[0], sr)
sf.write("output_custom_voice_2.wav", wavs[1], sr)
```
For `Qwen3-TTS-12Hz-1.7B/0.6B-CustomVoice` models, the supported speaker list and speaker descriptions are provided below. We recommend using each speaker’s native language for the best quality. Of course, each speaker can speak any language supported by the model.
| Speaker | Voice Description | Native language |
| --- | --- | --- |
| Vivian | Bright, slightly edgy young female voice. | Chinese |
| Serena | Warm, gentle young female voice. | Chinese |
| Uncle_Fu | Seasoned male voice with a low, mellow timbre. | Chinese |
| Dylan | Youthful Beijing male voice with a clear, natural timbre. | Chinese (Beijing Dialect) |
| Eric | Lively Chengdu male voice with a slightly husky brightness. | Chinese (Sichuan Dialect) |
| Ryan | Dynamic male voice with strong rhythmic drive. | English |
| Aiden | Sunny American male voice with a clear midrange. | English |
| Ono_Anna | Playful Japanese female voice with a light, nimble timbre. | Japanese |
| Sohee | Warm Korean female voice with rich emotion. | Korean |
#### Voice Design
For the voice design model (`Qwen3-TTS-12Hz-1.7B-VoiceDesign`), you can use `generate_voice_design` to provide the target text and a natural-language `instruct` description.
```python
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
model = Qwen3TTSModel.from_pretrained(
"Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
device_map="cuda:0",
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# single inference
wavs, sr = model.generate_voice_design(
text="哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
language="Chinese",
instruct="体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
)
sf.write("output_voice_design.wav", wavs[0], sr)
# batch inference
wavs, sr = model.generate_voice_design(
text=[
"哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"It's in the top drawer... wait, it's empty? No way, that's impossible! I'm sure I put it there!"
],
language=["Chinese", "English"],
instruct=[
"体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
"Speak in an incredulous tone, but with a hint of panic beginning to creep into your voice."
]
)
sf.write("output_voice_design_1.wav", wavs[0], sr)
sf.write("output_voice_design_2.wav", wavs[1], sr)
```
#### Voice Clone
For the voice clone model (`Qwen3-TTS-12Hz-1.7B/0.6B-Base`), to clone a voice and synthesize new content, you just need to provide a reference audio clip (`ref_audio`) along with its transcript (`ref_text`). `ref_audio` can be a local file path, a URL, a base64 string, or a `(numpy_array, sample_rate)` tuple. If you set `x_vector_only_mode=True`, only the speaker embedding is used so `ref_text` is not required, but cloning quality may be reduced.
```python
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
model = Qwen3TTSModel.from_pretrained(
"Qwen/Qwen3-TTS-12Hz-1.7B-Base",
device_map="cuda:0",
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
ref_audio = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/clone.wav"
ref_text = "Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And thanks to you."
wavs, sr = model.generate_voice_clone(
text="I am solving the equation: x = [-b ± √(b²-4ac)] / 2a? Nobody can — it's a disaster (◍•͈⌔•͈◍), very sad!",
language="English",
ref_audio=ref_audio,
ref_text=ref_text,
)
sf.write("output_voice_clone.wav", wavs[0], sr)
```
If you need to reuse the same reference prompt across multiple generations (to avoid recomputing prompt features), build it once with `create_voice_clone_prompt` and pass it via `voice_clone_prompt`.
```python
prompt_items = model.create_voice_clone_prompt(
ref_audio=ref_audio,
ref_text=ref_text,
x_vector_only_mode=False,
)
wavs, sr = model.generate_voice_clone(
text=["Sentence A.", "Sentence B."],
language=["English", "English"],
voice_clone_prompt=prompt_items,
)
sf.write("output_voice_clone_1.wav", wavs[0], sr)
sf.write("output_voice_clone_2.wav", wavs[1], sr)
```
For more examples of reusable voice clone prompts, batch cloning, and batch inference, please refer to the [example codes](https://github.com/QwenLM/Qwen3-TTS/blob/main/examples/test_model_12hz_base.py). With those examples and the `generate_voice_clone` function description, you can explore more advanced usage patterns.
#### Voice Design then Clone
If you want a designed voice that you can reuse like a cloned speaker, a practical workflow is: (1) use the **VoiceDesign** model to synthesize a short reference clip that matches your target persona, (2) feed that clip into `create_voice_clone_prompt` to build a reusable prompt, and then (3) call `generate_voice_clone` with `voice_clone_prompt` to generate new content without re-extracting features every time. This is especially useful when you want a consistent character voice across many lines.
```python
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
# create a reference audio in the target style using the VoiceDesign model
design_model = Qwen3TTSModel.from_pretrained(
"Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
device_map="cuda:0",
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
ref_text = "H-hey! You dropped your... uh... calculus notebook? I mean, I think it's yours? Maybe?"
ref_instruct = "Male, 17 years old, tenor range, gaining confidence - deeper breath support now, though vowels still tighten when nervous"
ref_wavs, sr = design_model.generate_voice_design(
text=ref_text,
language="English",
instruct=ref_instruct
)
sf.write("voice_design_reference.wav", ref_wavs[0], sr)
# build a reusable clone prompt from the voice design reference
clone_model = Qwen3TTSModel.from_pretrained(
"Qwen/Qwen3-TTS-12Hz-1.7B-Base",
device_map="cuda:0",
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
voice_clone_prompt = clone_model.create_voice_clone_prompt(
ref_audio=(ref_wavs[0], sr), # or "voice_design_reference.wav"
ref_text=ref_text,
)
sentences = [
"No problem! I actually... kinda finished those already? If you want to compare answers or something...",
"What? No! I mean yes but not like... I just think you're... your titration technique is really precise!",
]
# reuse it for multiple single calls
wavs, sr = clone_model.generate_voice_clone(
text=sentences[0],
language="English",
voice_clone_prompt=voice_clone_prompt,
)
sf.write("clone_single_1.wav", wavs[0], sr)
wavs, sr = clone_model.generate_voice_clone(
text=sentences[1],
language="English",
voice_clone_prompt=voice_clone_prompt,
)
sf.write("clone_single_2.wav", wavs[0], sr)
# or batch generate in one call
wavs, sr = clone_model.generate_voice_clone(
text=sentences,
language=["English", "English"],
voice_clone_prompt=voice_clone_prompt,
)
for i, w in enumerate(wavs):
sf.write(f"clone_batch_{i}.wav", w, sr)
```
#### Tokenizer Encode and Decode
If you only want to encode and decode audio for transport or training and so on, `Qwen3TTSTokenizer` supports encode/decode with paths, URLs, numpy waveforms, and dict/list payloads, for example:
```python
import soundfile as sf
from qwen_tts import Qwen3TTSTokenizer
tokenizer = Qwen3TTSTokenizer.from_pretrained(
"Qwen/Qwen3-TTS-Tokenizer-12Hz",
device_map="cuda:0",
)
enc = tokenizer.encode("https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/tokenizer_demo_1.wav")
wavs, sr = tokenizer.decode(enc)
sf.write("decode_output.wav", wavs[0], sr)
```
For more tokenizer examples (including different input formats and batch usage), please refer to the [example codes](https://github.com/QwenLM/Qwen3-TTS/blob/main/examples/test_tokenizer_12hz.py). With those examples and the description for `Qwen3TTSTokenizer`, you can explore more advanced usage patterns.
### Launch Local Web UI Demo
To launch the Qwen3-TTS web ui demo, simply install the `qwen-tts` package and run `qwen-tts-demo`. Use the command below for help:
```bash
qwen-tts-demo --help
```
To launch the demo, you can use the following commands:
```bash
# CustomVoice model
qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --ip 0.0.0.0 --port 8000
# VoiceDesign model
qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign --ip 0.0.0.0 --port 8000
# Base model
qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-Base --ip 0.0.0.0 --port 8000
```
And then open `http://<your-ip>:8000`, or access it via port forwarding in tools like VS Code.
#### Base Model HTTPS Notes
To avoid browser microphone permission issues after deploying the server, for Base model deployments, it is recommended/required to run the gradio service over **HTTPS** (especially when accessed remotely or behind modern browsers/gateways). Use `--ssl-certfile` and `--ssl-keyfile` to enable HTTPS. First we need to generate a private key and a self-signed cert (valid for 365 days):
```bash
openssl req -x509 -newkey rsa:2048 \
-keyout key.pem -out cert.pem \
-days 365 -nodes \
-subj "/CN=localhost"
```
Then run the demo with HTTPS:
```bash
qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-Base \
--ip 0.0.0.0 --port 8000 \
--ssl-certfile cert.pem \
--ssl-keyfile key.pem \
--no-ssl-verify
```
And open `https://<your-ip>:8000` to experience it. If your browser shows a warning, it’s expected for self-signed certificates. For production, use a real certificate.
### DashScope API Usage
To further explore Qwen3-TTS, we encourage you to try our DashScope API for a faster and more efficient experience. For detailed API information and documentation, please refer to the following:
| API Description | API Documentation (Mainland China) | API Documentation (International) |
|------------------|-----------------------------------|------------------------------------|
| Real-time API for Qwen3-TTS of custom voice model. | [https://help.aliyun.com/zh/model-studio/qwen-tts-realtime](https://help.aliyun.com/zh/model-studio/qwen-tts-realtime) | [https://www.alibabacloud.com/help/en/model-studio/qwen-tts-realtime](https://www.alibabacloud.com/help/en/model-studio/qwen-tts-realtime) |
| Real-time API for Qwen3-TTS of voice clone model. | [https://help.aliyun.com/zh/model-studio/qwen-tts-voice-cloning](https://help.aliyun.com/zh/model-studio/qwen-tts-voice-cloning) | [https://www.alibabacloud.com/help/en/model-studio/qwen-tts-voice-cloning](https://www.alibabacloud.com/help/en/model-studio/qwen-tts-voice-cloning) |
| Real-time API for Qwen3-TTS of voice design model. | [https://help.aliyun.com/zh/model-studio/qwen-tts-voice-design](https://help.aliyun.com/zh/model-studio/qwen-tts-voice-design) | [https://www.alibabacloud.com/help/en/model-studio/qwen-tts-voice-design](https://www.alibabacloud.com/help/en/model-studio/qwen-tts-voice-design) |
## vLLM Usage
vLLM officially provides day-0 support for Qwen3-TTS! Welcome to use vLLM-Omni for Qwen3-TTS deployment and inference. For installation and more details, please check [vLLM-Omni official documentation](https://docs.vllm.ai/projects/vllm-omni/en/latest/getting_started/quickstart/#installation). Now only offline inference is supported. Online serving will be supported later, and vLLM-Omni will continue to offer support and optimization for Qwen3-TTS in areas such as inference speed and streaming capabilities.
### Offline Inference
You can use vLLM-Omni to inference Qwen3-TTS locally, we provide examples in [vLLM-Omni repo](https://github.com/vllm-project/vllm-omni/tree/main/examples/offline_inference/qwen3_tts) which can generate audio output:
```bash
# git clone https://github.com/vllm-project/vllm-omni.git
# cd vllm-omni/examples/offline_inference/qwen3_tts
# Run a single sample with CustomVoice task
python end2end.py --query-type CustomVoice
# Batch sample (multiple prompts in one run) with CustomVoice task:
python end2end.py --query-type CustomVoice --use-batch-sample
# Run a single sample with VoiceDesign task
python end2end.py --query-type VoiceDesign
# Batch sample (multiple prompts in one run) with VoiceDesign task:
python end2end.py --query-type VoiceDesign --use-batch-sample
# Run a single sample with Base task in icl mode-tag
python end2end.py --query-type Base --mode-tag icl
```
## Evaluation
During evaluation, we ran inference for all models with `dtype=torch.bfloat16` and set `max_new_tokens=2048`. All other sampling parameters used the defaults from the checkpoint’s `generate_config.json`. For the Seed-Test and InstructTTS-Eval test sets, we set `language="auto"`, while for all other test sets we explicitly passed the corresponding `language`. The detailed results are shown below.
<details>
<summary>Speech Generation Benchmarks</summary>
*Zero-shot speech generation on the Seed-TTS test set. Performance is measured by Word Error Rate (WER, ↓), where lower is better.*
<table>
<thead>
<tr>
<th style="text-align: center;">Datasets</th>
<th style="text-align: left;">Model</th>
<th colspan="2" style="text-align: center;">Performance</th>
</tr>
<tr style="border-bottom: 1px solid #ddd; border-top: 1px solid #ddd;">
<td colspan="4" style="text-align: center;"><em>Content Consistency</em></td>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="14" style="text-align: center; vertical-align: middle;">SEED<br><em>test-zh</em> | <em>test-en</em></td>
<td style="text-align: left;">Seed-TTS (Anastassiou et al., 2024)</td>
<td style="text-align: center;">1.12</td>
<td style="text-align: center;">2.25</td>
</tr>
<tr>
<td style="text-align: left;">MaskGCT (Wang et al., 2024)</td>
<td style="text-align: center;">2.27</td>
<td style="text-align: center;">2.62</td>
</tr>
<tr>
<td style="text-align: left;">E2 TTS (Eskimez et al., 2024)</td>
<td style="text-align: center;">1.97</td>
<td style="text-align: center;">2.19</td>
</tr>
<tr>
<td style="text-align: left;">F5-TTS (Chen et al., 2024)</td>
<td style="text-align: center;">1.56</td>
<td style="text-align: center;">1.83</td>
</tr>
<tr>
<td style="text-align: left;">Spark TTS (Wang et al., 2025)</td>
<td style="text-align: center;">1.20</td>
<td style="text-align: center;">1.98</td>
</tr>
<tr>
<td style="text-align: left;">Llasa-8B (Ye et al., 2025b)</td>
<td style="text-align: center;">1.59</td>
<td style="text-align: center;">2.97</td>
</tr>
<tr>
<td style="text-align: left;">KALL-E (Xia et al., 2024)</td>
<td style="text-align: center;">0.96</td>
<td style="text-align: center;">1.94</td>
</tr>
<tr>
<td style="text-align: left;">FireRedTTS 2 (Xie et al., 2025)</td>
<td style="text-align: center;">1.14</td>
<td style="text-align: center;">1.95</td>
</tr>
<tr>
<td style="text-align: left;">CosyVoice 3 (Du et al., 2025)</td>
<td style="text-align: center;"><strong>0.71</strong></td>
<td style="text-align: center;">1.45</td>
</tr>
<tr>
<td style="text-align: left;">MiniMax-Speech (Zhang et al., 2025a)</td>
<td style="text-align: center;">0.83</td>
<td style="text-align: center;">1.65</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-25Hz-0.6B-Base</td>
<td style="text-align: center;">1.18</td>
<td style="text-align: center;">1.64</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-25Hz-1.7B-Base</td>
<td style="text-align: center;">1.10</td>
<td style="text-align: center;">1.49</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-12Hz-0.6B-Base</td>
<td style="text-align: center;">0.92</td>
<td style="text-align: center;">1.32</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-12Hz-1.7B-Base</td>
<td style="text-align: center;">0.77</td>
<td style="text-align: center;"><strong>1.24</strong></td>
</tr>
</tbody>
</table>
<br>
*Multilingual speech generation on the TTS multilingual test set. Performance is measured by Word Error Rate (WER, ↓) for content consistency and Cosine Similarity (SIM, ↑) for speaker similarity.*
<table>
<thead>
<tr>
<th rowspan="2" style="text-align: left; vertical-align: bottom;">Language</th>
<th colspan="2" style="text-align: center;">Qwen3-TTS-25Hz</th>
<th colspan="2" style="text-align: center;">Qwen3-TTS-12Hz</th>
<th rowspan="2" style="text-align: center; vertical-align: bottom;">MiniMax</th>
<th rowspan="2" style="text-align: center; vertical-align: bottom;">ElevenLabs</th>
</tr>
<tr>
<th style="text-align: center;">0.6B-Base</th>
<th style="text-align: center;">1.7B-Base</th>
<th style="text-align: center;">0.6B-Base</th>
<th style="text-align: center;">1.7B-Base</th>
</tr>
</thead>
<tbody>
<tr>
<td colspan="7" style="text-align: center; border-top: 1px solid #ddd; border-bottom: 1px solid #ddd;"><em>Content Consistency</em></td>
</tr>
<tr>
<td style="text-align: left;">Chinese</td>
<td style="text-align: center;">1.108</td>
<td style="text-align: center;"><strong>0.777</strong></td>
<td style="text-align: center;">1.145</td>
<td style="text-align: center;">0.928</td>
<td style="text-align: center;">2.252</td>
<td style="text-align: center;">16.026</td>
</tr>
<tr>
<td style="text-align: left;">English</td>
<td style="text-align: center;">1.048</td>
<td style="text-align: center;">1.014</td>
<td style="text-align: center;"><strong>0.836</strong></td>
<td style="text-align: center;">0.934</td>
<td style="text-align: center;">2.164</td>
<td style="text-align: center;">2.339</td>
</tr>
<tr>
<td style="text-align: left;">German</td>
<td style="text-align: center;">1.501</td>
<td style="text-align: center;">0.960</td>
<td style="text-align: center;">1.089</td>
<td style="text-align: center;">1.235</td>
<td style="text-align: center;">1.906</td>
<td style="text-align: center;"><strong>0.572</strong></td>
</tr>
<tr>
<td style="text-align: left;">Italian</td>
<td style="text-align: center;">1.169</td>
<td style="text-align: center;">1.105</td>
<td style="text-align: center;">1.534</td>
<td style="text-align: center;"><strong>0.948</strong></td>
<td style="text-align: center;">1.543</td>
<td style="text-align: center;">1.743</td>
</tr>
<tr>
<td style="text-align: left;">Portuguese</td>
<td style="text-align: center;">2.046</td>
<td style="text-align: center;">1.778</td>
<td style="text-align: center;">2.254</td>
<td style="text-align: center;">1.526</td>
<td style="text-align: center;">1.877</td>
<td style="text-align: center;"><strong>1.331</strong></td>
</tr>
<tr>
<td style="text-align: left;">Spanish</td>
<td style="text-align: center;">2.031</td>
<td style="text-align: center;">1.491</td>
<td style="text-align: center;">1.491</td>
<td style="text-align: center;">1.126</td>
<td style="text-align: center;"><strong>1.029</strong></td>
<td style="text-align: center;">1.084</td>
</tr>
<tr>
<td style="text-align: left;">Japanese</td>
<td style="text-align: center;">4.189</td>
<td style="text-align: center;">5.121</td>
<td style="text-align: center;">6.404</td>
<td style="text-align: center;">3.823</td>
<td style="text-align: center;"><strong>3.519</strong></td>
<td style="text-align: center;">10.646</td>
</tr>
<tr>
<td style="text-align: left;">Korean</td>
<td style="text-align: center;">2.852</td>
<td style="text-align: center;">2.631</td>
<td style="text-align: center;"><strong>1.741</strong></td>
<td style="text-align: center;">1.755</td>
<td style="text-align: center;">1.747</td>
<td style="text-align: center;">1.865</td>
</tr>
<tr>
<td style="text-align: left;">French</td>
<td style="text-align: center;">2.852</td>
<td style="text-align: center;"><strong>2.631</strong></td>
<td style="text-align: center;">2.931</td>
<td style="text-align: center;">2.858</td>
<td style="text-align: center;">4.099</td>
<td style="text-align: center;">5.216</td>
</tr>
<tr>
<td style="text-align: left;">Russian</td>
<td style="text-align: center;">5.957</td>
<td style="text-align: center;">4.535</td>
<td style="text-align: center;">4.458</td>
<td style="text-align: center;"><strong>3.212</strong></td>
<td style="text-align: center;">4.281</td>
<td style="text-align: center;">3.878</td>
</tr>
<tr style="border-top: 1px solid #ddd;">
<td colspan="7" style="text-align: center; border-bottom: 1px solid #ddd;"><em>Speaker Similarity</em></td>
</tr>
<tr>
<td style="text-align: left;">Chinese</td>
<td style="text-align: center;">0.797</td>
<td style="text-align: center;">0.796</td>
<td style="text-align: center;"><strong>0.811</strong></td>
<td style="text-align: center;">0.799</td>
<td style="text-align: center;">0.780</td>
<td style="text-align: center;">0.677</td>
</tr>
<tr>
<td style="text-align: left;">English</td>
<td style="text-align: center;">0.811</td>
<td style="text-align: center;">0.815</td>
<td style="text-align: center;"><strong>0.829</strong></td>
<td style="text-align: center;">0.775</td>
<td style="text-align: center;">0.756</td>
<td style="text-align: center;">0.613</td>
</tr>
<tr>
<td style="text-align: left;">German</td>
<td style="text-align: center;">0.749</td>
<td style="text-align: center;">0.737</td>
<td style="text-align: center;">0.769</td>
<td style="text-align: center;"><strong>0.775</strong></td>
<td style="text-align: center;">0.733</td>
<td style="text-align: center;">0.614</td>
</tr>
<tr>
<td style="text-align: left;">Italian</td>
<td style="text-align: center;">0.722</td>
<td style="text-align: center;">0.718</td>
<td style="text-align: center;">0.792</td>
<td style="text-align: center;"><strong>0.817</strong></td>
<td style="text-align: center;">0.699</td>
<td style="text-align: center;">0.579</td>
</tr>
<tr>
<td style="text-align: left;">Portuguese</td>
<td style="text-align: center;">0.790</td>
<td style="text-align: center;">0.783</td>
<td style="text-align: center;">0.794</td>
<td style="text-align: center;"><strong>0.817</strong></td>
<td style="text-align: center;">0.805</td>
<td style="text-align: center;">0.711</td>
</tr>
<tr>
<td style="text-align: left;">Spanish</td>
<td style="text-align: center;">0.732</td>
<td style="text-align: center;">0.731</td>
<td style="text-align: center;">0.812</td>
<td style="text-align: center;"><strong>0.814</strong></td>
<td style="text-align: center;">0.762</td>
<td style="text-align: center;">0.615</td>
</tr>
<tr>
<td style="text-align: left;">Japanese</td>
<td style="text-align: center;"><strong>0.810</strong></td>
<td style="text-align: center;">0.807</td>
<td style="text-align: center;">0.798</td>
<td style="text-align: center;">0.788</td>
<td style="text-align: center;">0.776</td>
<td style="text-align: center;">0.738</td>
</tr>
<tr>
<td style="text-align: left;">Korean</td>
<td style="text-align: center;"><strong>0.824</strong></td>
<td style="text-align: center;">0.814</td>
<td style="text-align: center;">0.812</td>
<td style="text-align: center;">0.799</td>
<td style="text-align: center;">0.779</td>
<td style="text-align: center;">0.700</td>
</tr>
<tr>
<td style="text-align: left;">French</td>
<td style="text-align: center;">0.698</td>
<td style="text-align: center;">0.703</td>
<td style="text-align: center;">0.700</td>
<td style="text-align: center;"><strong>0.714</strong></td>
<td style="text-align: center;">0.628</td>
<td style="text-align: center;">0.535</td>
</tr>
<tr>
<td style="text-align: left;">Russian</td>
<td style="text-align: center;">0.734</td>
<td style="text-align: center;">0.744</td>
<td style="text-align: center;">0.781</td>
<td style="text-align: center;"><strong>0.792</strong></td>
<td style="text-align: center;">0.761</td>
<td style="text-align: center;">0.676</td>
</tr>
</tbody>
</table>
<br>
*Cross-lingual speech generation on the Cross-Lingual benchmark. Performance is measured by Mixed Error Rate (WER for English, CER for others, ↓).*
<table>
<thead>
<tr>
<th style="text-align: left;">Task</th>
<th style="text-align: center;">Qwen3-TTS-25Hz-1.7B-Base</th>
<th style="text-align: center;">Qwen3-TTS-12Hz-1.7B-Base</th>
<th style="text-align: center;">CosyVoice3</th>
<th style="text-align: center;">CosyVoice2</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align: left;">en-to-zh</td>
<td style="text-align: center;">5.66</td>
<td style="text-align: center;"><strong>4.77</strong></td>
<td style="text-align: center;">5.09</td>
<td style="text-align: center;">13.5</td>
</tr>
<tr>
<td style="text-align: left;">ja-to-zh</td>
<td style="text-align: center;">3.92</td>
<td style="text-align: center;">3.43</td>
<td style="text-align: center;"><strong>3.05</strong></td>
<td style="text-align: center;">48.1</td>
</tr>
<tr>
<td style="text-align: left;">ko-to-zh</td>
<td style="text-align: center;">1.14</td>
<td style="text-align: center;">1.08</td>
<td style="text-align: center;"><strong>1.06</strong></td>
<td style="text-align: center;">7.70</td>
</tr>
<tr style="border-top: 1px solid #ddd;">
<td style="text-align: left;">zh-to-en</td>
<td style="text-align: center;">2.91</td>
<td style="text-align: center;"><strong>2.77</strong></td>
<td style="text-align: center;">2.98</td>
<td style="text-align: center;">6.47</td>
</tr>
<tr>
<td style="text-align: left;">ja-to-en</td>
<td style="text-align: center;">3.95</td>
<td style="text-align: center;"><strong>3.04</strong></td>
<td style="text-align: center;">4.20</td>
<td style="text-align: center;">17.1</td>
</tr>
<tr>
<td style="text-align: left;">ko-to-en</td>
<td style="text-align: center;">3.48</td>
<td style="text-align: center;"><strong>3.09</strong></td>
<td style="text-align: center;">4.19</td>
<td style="text-align: center;">11.2</td>
</tr>
<tr style="border-top: 1px solid #ddd;">
<td style="text-align: left;">zh-to-ja</td>
<td style="text-align: center;">9.29</td>
<td style="text-align: center;">8.40</td>
<td style="text-align: center;"><strong>7.08</strong></td>
<td style="text-align: center;">13.1</td>
</tr>
<tr>
<td style="text-align: left;">en-to-ja</td>
<td style="text-align: center;">7.74</td>
<td style="text-align: center;">7.21</td>
<td style="text-align: center;"><strong>6.80</strong></td>
<td style="text-align: center;">14.9</td>
</tr>
<tr>
<td style="text-align: left;">ko-to-ja</td>
<td style="text-align: center;">4.17</td>
<td style="text-align: center;"><strong>3.67</strong></td>
<td style="text-align: center;">3.93</td>
<td style="text-align: center;">5.86</td>
</tr>
<tr style="border-top: 1px solid #ddd;">
<td style="text-align: left;">zh-to-ko</td>
<td style="text-align: center;">8.12</td>
<td style="text-align: center;"><strong>4.82</strong></td>
<td style="text-align: center;">14.4</td>
<td style="text-align: center;">24.8</td>
</tr>
<tr>
<td style="text-align: left;">en-to-ko</td>
<td style="text-align: center;">6.83</td>
<td style="text-align: center;"><strong>5.14</strong></td>
<td style="text-align: center;">5.87</td>
<td style="text-align: center;">21.9</td>
</tr>
<tr>
<td style="text-align: left;">ja-to-ko</td>
<td style="text-align: center;">6.86</td>
<td style="text-align: center;"><strong>5.59</strong></td>
<td style="text-align: center;">7.92</td>
<td style="text-align: center;">21.5</td>
</tr>
</tbody>
</table>
<br>
*Controllable speech generation on InstructTTSEval. Performance is measured by Attribute Perception and Synthesis accuracy (APS), Description-Speech Consistency (DSD), and Response Precision (RP).*
<table>
<thead>
<tr>
<th rowspan="2" style="text-align: left; vertical-align: bottom;">Type</th>
<th rowspan="2" style="text-align: left; vertical-align: bottom;">Model</th>
<th colspan="3" style="text-align: center;">InstructTTSEval-ZH</th>
<th colspan="3" style="text-align: center;">InstructTTSEval-EN</th>
</tr>
<tr>
<th style="text-align: center;">APS (↑)</th>
<th style="text-align: center;">DSD (↑)</th>
<th style="text-align: center;">RP (↑)</th>
<th style="text-align: center;">APS (↑)</th>
<th style="text-align: center;">DSD (↑)</th>
<th style="text-align: center;">RP (↑)</th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="5" style="text-align: left; vertical-align: middle;"><em>Target<br>Speaker</em></td>
<td style="text-align: left;">Gemini-flash</td>
<td style="text-align: center;">88.2</td>
<td style="text-align: center;"><strong>90.9</strong></td>
<td style="text-align: center;"><strong>77.3</strong></td>
<td style="text-align: center;"><strong>92.3</strong></td>
<td style="text-align: center;"><strong>93.8</strong></td>
<td style="text-align: center;"><strong>80.1</strong></td>
</tr>
<tr>
<td style="text-align: left;">Gemini-pro</td>
<td style="text-align: center;"><strong>89.0</strong></td>
<td style="text-align: center;">90.1</td>
<td style="text-align: center;">75.5</td>
<td style="text-align: center;">87.6</td>
<td style="text-align: center;">86.0</td>
<td style="text-align: center;">67.2</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3TTS-25Hz-1.7B-CustomVoice</td>
<td style="text-align: center;">83.1</td>
<td style="text-align: center;">75.0</td>
<td style="text-align: center;">63.0</td>
<td style="text-align: center;">79.0</td>
<td style="text-align: center;">82.8</td>
<td style="text-align: center;">69.3</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3TTS-12Hz-1.7B-CustomVoice</td>
<td style="text-align: center;">83.0</td>
<td style="text-align: center;">77.8</td>
<td style="text-align: center;">61.2</td>
<td style="text-align: center;">77.3</td>
<td style="text-align: center;">77.1</td>
<td style="text-align: center;">63.7</td>
</tr>
<tr>
<td style="text-align: left;">GPT-4o-mini-tts</td>
<td style="text-align: center;">54.9</td>
<td style="text-align: center;">52.3</td>
<td style="text-align: center;">46.0</td>
<td style="text-align: center;">76.4</td>
<td style="text-align: center;">74.3</td>
<td style="text-align: center;">54.8</td>
</tr>
<tr style="border-top: 1px solid #ddd;">
<td rowspan="9" style="text-align: left; vertical-align: middle;"><em>Voice<br>Design</em></td>
<td style="text-align: left;">Qwen3TTS-12Hz-1.7B-VD</td>
<td style="text-align: center;"><strong>85.2</strong></td>
<td style="text-align: center;"><strong>81.1</strong></td>
<td style="text-align: center;"><strong>65.1</strong></td>
<td style="text-align: center;">82.9</td>
<td style="text-align: center;"><strong>82.4</strong></td>
<td style="text-align: center;"><strong>68.4</strong></td>
</tr>
<tr>
<td style="text-align: left;">Mimo-Audio-7B-Instruct (Zhang et al., 2025b)</td>
<td style="text-align: center;">75.7</td>
<td style="text-align: center;">74.3</td>
<td style="text-align: center;">61.5</td>
<td style="text-align: center;">80.6</td>
<td style="text-align: center;">77.6</td>
<td style="text-align: center;">59.5</td>
</tr>
<tr>
<td style="text-align: left;">VoiceSculptor (Hu et al., 2026)</td>
<td style="text-align: center;">75.7</td>
<td style="text-align: center;">64.7</td>
<td style="text-align: center;">61.5</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
</tr>
<tr>
<td style="text-align: left;">Hume</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;"><strong>83.0</strong></td>
<td style="text-align: center;">75.3</td>
<td style="text-align: center;">54.3</td>
</tr>
<tr>
<td style="text-align: left;">VoxInstruct (Zhou et al., 2024)</td>
<td style="text-align: center;">47.5</td>
<td style="text-align: center;">52.3</td>
<td style="text-align: center;">42.6</td>
<td style="text-align: center;">54.9</td>
<td style="text-align: center;">57.0</td>
<td style="text-align: center;">39.3</td>
</tr>
<tr>
<td style="text-align: left;">Parler-tts-mini (Lyth & King, 2024)</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">63.4</td>
<td style="text-align: center;">48.7</td>
<td style="text-align: center;">28.6</td>
</tr>
<tr>
<td style="text-align: left;">Parler-tts-large (Lyth & King, 2024)</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">60.0</td>
<td style="text-align: center;">45.9</td>
<td style="text-align: center;">31.2</td>
</tr>
<tr>
<td style="text-align: left;">PromptTTS (Guo et al., 2023)</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">64.3</td>
<td style="text-align: center;">47.2</td>
<td style="text-align: center;">31.4</td>
</tr>
<tr>
<td style="text-align: left;">PromptStyle (Liu et al., 2023)</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">57.4</td>
<td style="text-align: center;">46.4</td>
<td style="text-align: center;">30.9</td>
</tr>
</tbody>
</table>
<br>
*Target-Speaker Multilingual Speech Generation on the TTS multilingual test set. Performance is measured by Word Error Rate (WER, ↓).*
<table>
<thead>
<tr>
<th rowspan="2" style="text-align: left; vertical-align: bottom;">Language</th>
<th colspan="2" style="text-align: center;">Qwen3-TTS-25Hz</th>
<th colspan="2" style="text-align: center;">Qwen3-TTS-12Hz</th>
<th rowspan="2" style="text-align: center; vertical-align: bottom;">GPT-4o-Audio<br>Preview</th>
</tr>
<tr>
<th style="text-align: center;">0.6B-CustomVoice</th>
<th style="text-align: center;">1.7B-CustomVoice</th>
<th style="text-align: center;">0.6B-CustomVoice</th>
<th style="text-align: center;">1.7B-CustomVoice</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align: left;">Chinese</td>
<td style="text-align: center;">0.874</td>
<td style="text-align: center;"><strong>0.708</strong></td>
<td style="text-align: center;">0.944</td>
<td style="text-align: center;">0.903</td>
<td style="text-align: center;">3.519</td>
</tr>
<tr>
<td style="text-align: left;">English</td>
<td style="text-align: center;">1.332</td>
<td style="text-align: center;">0.936</td>
<td style="text-align: center;">1.188</td>
<td style="text-align: center;"><strong>0.899</strong></td>
<td style="text-align: center;">2.197</td>
</tr>
<tr>
<td style="text-align: left;">German</td>
<td style="text-align: center;">0.990</td>
<td style="text-align: center;"><strong>0.634</strong></td>
<td style="text-align: center;">2.722</td>
<td style="text-align: center;">1.057</td>
<td style="text-align: center;">1.161</td>
</tr>
<tr>
<td style="text-align: left;">Italian</td>
<td style="text-align: center;">1.861</td>
<td style="text-align: center;">1.271</td>
<td style="text-align: center;">2.545</td>
<td style="text-align: center;">1.362</td>
<td style="text-align: center;"><strong>1.194</strong></td>
</tr>
<tr>
<td style="text-align: left;">Portuguese</td>
<td style="text-align: center;">1.728</td>
<td style="text-align: center;">1.854</td>
<td style="text-align: center;">3.219</td>
<td style="text-align: center;">2.681</td>
<td style="text-align: center;"><strong>1.504</strong></td>
</tr>
<tr>
<td style="text-align: left;">Spanish</td>
<td style="text-align: center;">1.309</td>
<td style="text-align: center;">1.284</td>
<td style="text-align: center;"><strong>1.154</strong></td>
<td style="text-align: center;">1.330</td>
<td style="text-align: center;">4.000</td>
</tr>
<tr>
<td style="text-align: left;">Japanese</td>
<td style="text-align: center;"><strong>3.875</strong></td>
<td style="text-align: center;">4.518</td>
<td style="text-align: center;">6.877</td>
<td style="text-align: center;">4.924</td>
<td style="text-align: center;">5.001</td>
</tr>
<tr>
<td style="text-align: left;">Korean</td>
<td style="text-align: center;">2.202</td>
<td style="text-align: center;">2.274</td>
<td style="text-align: center;">3.053</td>
<td style="text-align: center;"><strong>1.741</strong></td>
<td style="text-align: center;">2.763</td>
</tr>
<tr>
<td style="text-align: left;">French</td>
<td style="text-align: center;">3.865</td>
<td style="text-align: center;"><strong>3.080</strong></td>
<td style="text-align: center;">3.841</td>
<td style="text-align: center;">3.781</td>
<td style="text-align: center;">3.605</td>
</tr>
<tr>
<td style="text-align: left;">Russian</td>
<td style="text-align: center;">6.529</td>
<td style="text-align: center;"><strong>4.444</strong></td>
<td style="text-align: center;">5.809</td>
<td style="text-align: center;">4.734</td>
<td style="text-align: center;">5.250</td>
</tr>
</tbody>
</table>
<br>
*Long speech generation results. Performance is measured by Word Error Rate (WER, ↓).*
<table>
<thead>
<tr>
<th style="text-align: center;">Datasets</th>
<th style="text-align: left;">Model</th>
<th colspan="2" style="text-align: center;">Performance</th>
</tr>
<tr style="border-bottom: 1px solid #ddd; border-top: 1px solid #ddd;">
<td colspan="4" style="text-align: center;"><em>Content Consistency</em></td>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="5" style="text-align: center; vertical-align: middle;"><em>long-zh</em> | <em>long-en</em></td>
<td style="text-align: left;">Higgs-Audio-v2 (chunk) (Boson AI, 2025)</td>
<td style="text-align: center;">5.505</td>
<td style="text-align: center;">6.917</td>
</tr>
<tr>
<td style="text-align: left;">VibeVoice (Peng et al., 2025)</td>
<td style="text-align: center;">22.619</td>
<td style="text-align: center;">1.780</td>
</tr>
<tr>
<td style="text-align: left;">VoxCPM (Zhou et al., 2025)</td>
<td style="text-align: center;">4.835</td>
<td style="text-align: center;">7.474</td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-25Hz-1.7B-CustomVoice</td>
<td style="text-align: center;"><strong>1.517</strong></td>
<td style="text-align: center;"><strong>1.225</strong></td>
</tr>
<tr>
<td style="text-align: left;">Qwen3-TTS-12Hz-1.7B-CustomVoice</td>
<td style="text-align: center;">2.356</td>
<td style="text-align: center;">2.812</td>
</tr>
</tbody>
</table>
</details>
<details>
<summary>Speech Tokenizer Benchmarks</summary>
*Comparison between different supervised semantic speech tokenizers on ASR Task.*
<table>
<thead>
<tr>
<th style="text-align: left;">Model</th>
<th style="text-align: center;">Codebook Size</th>
<th style="text-align: center;">FPS</th>
<th style="text-align: center;">C.V. EN</th>
<th style="text-align: center;">C.V. CN</th>
<th style="text-align: center;">Fluers EN</th>
<th style="text-align: center;">Fluers CN</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align: left;">S3 Tokenizer(VQ) (Du et al., 2024a)</td>
<td style="text-align: center;">4096</td>
<td style="text-align: center;">50</td>
<td style="text-align: center;">12.06</td>
<td style="text-align: center;">15.38</td>
<td style="text-align: center;">-</td>
<td style="text-align: center;">-</td>
</tr>
<tr>
<td style="text-align: left;">S3 Tokenizer(VQ) (Du et al., 2024a)</td>
<td style="text-align: center;">4096</td>
<td style="text-align: center;">25</td>
<td style="text-align: center;">11.56</td>
<td style="text-align: center;">18.26</td>
<td style="text-align: center;">7.65</td>
<td style="text-align: center;">5.03</td>
</tr>
<tr>
<td style="text-align: left;">S3 Tokenizer(FSQ) (Du et al., 2024a)</td>
<td style="text-align: center;">6561</td>
<td style="text-align: center;">25</td>
<td style="text-align: center;">10.67</td>
<td style="text-align: center;"><strong>7.29</strong></td>
<td style="text-align: center;">6.58</td>
<td style="text-align: center;">4.43</td>
</tr>
<tr>
<td style="text-align: left;">Qwen-TTS-Tokenizer-25Hz (Stage 1)</td>
<td style="text-align: center;">32768</td>
<td style="text-align: center;">25</td>
<td style="text-align: center;"><strong>7.51</strong></td>
<td style="text-align: center;">10.73</td>
<td style="text-align: center;"><strong>3.07</strong></td>
<td style="text-align: center;"><strong>4.23</strong></td>
</tr>
<tr>
<td style="text-align: left;">Qwen-TTS-Tokenizer-25Hz (Stage 2)</td>
<td style="text-align: center;">32768</td>
<td style="text-align: center;">25</td>
<td style="text-align: center;">10.40</td>
<td style="text-align: center;">14.99</td>
<td style="text-align: center;">4.14</td>
<td style="text-align: center;">4.67</td>
</tr>
</tbody>
</table>
<br>
*Comparison between different semantic-related speech tokenizers.*
<table>
<thead>
<tr>
<th style="text-align: left;">Model</th>
<th style="text-align: center;">NQ</th>
<th style="text-align: center;">Codebook Size</th>
<th style="text-align: center;">FPS</th>
<th style="text-align: center;">PESQ_WB</th>
<th style="text-align: center;">PESQ_NB</th>
<th style="text-align: center;">STOI</th>
<th style="text-align: center;">UTMOS</th>
<th style="text-align: center;">SIM</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align: left;">SpeechTokenizer (Zhang et al., 2023a)</td>
<td style="text-align: center;">8</td>
<td style="text-align: center;">1024</td>
<td style="text-align: center;">50</td>
<td style="text-align: center;">2.60</td>
<td style="text-align: center;">3.05</td>
<td style="text-align: center;">0.92</td>
<td style="text-align: center;">3.90</td>
<td style="text-align: center;">0.85</td>
</tr>
<tr>
<td style="text-align: left;">X-codec (Ye et al., 2025a)</td>
<td style="text-align: center;">2</td>
<td style="text-align: center;">1024</td>
<td style="text-align: center;">50</td>
<td style="text-align: center;">2.68</td>
<td style="text-align: center;">3.27</td>
<td style="text-align: center;">0.86</td>
<td style="text-align: center;">4.11</td>
<td style="text-align: center;">0.84</td>
</tr>
<tr>
<td style="text-align: left;">X-codec 2 (Ye et al., 2025b)</td>
<td style="text-align: center;">1</td>
<td style="text-align: center;">65536</td>
<td style="text-align: center;">50</td>
<td style="text-align: center;">2.43</td>
<td style="text-align: center;">3.04</td>
<td style="text-align: center;">0.92</td>
<td style="text-align: center;">4.13</td>
<td style="text-align: center;">0.82</td>
</tr>
<tr>
<td style="text-align: left;">XY-Tokenizer (Gong et al., 2025)</td>
<td style="text-align: center;">8</td>
<td style="text-align: center;">1024</td>
<td style="text-align: center;">12.5</td>
<td style="text-align: center;">2.41</td>
<td style="text-align: center;">3.00</td>
<td style="text-align: center;">0.91</td>
<td style="text-align: center;">3.98</td>
<td style="text-align: center;">0.83</td>
</tr>
<tr>
<td style="text-align: left;">Mimi (Défossez et al., 2024)</td>
<td style="text-align: center;">16</td>
<td style="text-align: center;">2048</td>
<td style="text-align: center;">12.5</td>
<td style="text-align: center;">2.88</td>
<td style="text-align: center;">3.42</td>
<td style="text-align: center;">0.94</td>
<td style="text-align: center;">3.87</td>
<td style="text-align: center;">0.87</td>
</tr>
<tr>
<td style="text-align: left;">FireredTTS 2 Tokenizer (Xie et al., 2025)</td>
<td style="text-align: center;">16</td>
<td style="text-align: center;">2048</td>
<td style="text-align: center;">12.5</td>
<td style="text-align: center;">2.73</td>
<td style="text-align: center;">3.28</td>
<td style="text-align: center;">0.94</td>
<td style="text-align: center;">3.88</td>
<td style="text-align: center;">0.87</td>
</tr>
<tr>
<td style="text-align: left;">Qwen-TTS-Tokenizer-12Hz</td>
<td style="text-align: center;">16</td>
<td style="text-align: center;">2048</td>
<td style="text-align: center;">12.5</td>
<td style="text-align: center;"><strong>3.21</strong></td>
<td style="text-align: center;"><strong>3.68</strong></td>
<td style="text-align: center;"><strong>0.96</strong></td>
<td style="text-align: center;"><strong>4.16</strong></td>
<td style="text-align: center;"><strong>0.95</strong></td>
</tr>
</tbody>
</table>
</details>
## Citation
If you find our paper and code useful in your research, please consider giving a star :star: and citation :pencil: :)
```BibTeX
@article{Qwen3-TTS,
title={Qwen3-TTS Technical Report},
author={Hangrui Hu and Xinfa Zhu and Ting He and Dake Guo and Bin Zhang and Xiong Wang and Zhifang Guo and Ziyue Jiang and Hongkun Hao and Zishan Guo and Xinyu Zhang and Pei Zhang and Baosong Yang and Jin Xu and Jingren Zhou and Junyang Lin},
journal={arXiv preprint arXiv:2601.15621},
year={2026}
}
```
<br>
\ No newline at end of file
doc/iocn.png 0 → 100644
View file @ e3cdb305
File suppressed by a .gitattributes entry or the file's encoding is unsupported.
doc/qwen3-tts.png 0 → 100644
View file @ e3cdb305
File suppressed by a .gitattributes entry or the file's encoding is unsupported.
model.properties 0 → 100644
View file @ e3cdb305
# 模型唯一标识
modelCode=2047
# 模型名称
modelName=Qwen3-TTS_pytorch
# 模型描述
modelDescription=由通义开发的一系列强大的语音生成模型，支持声音克隆、声音设计、高质量拟人声生成和基于自然语言的语音控制。
# 运行过程
processType=推理
# 算法类别
appCategory=语音合成
# 框架类型
frameType=vllm
# 加速卡类型
accelerateType=K100AI
output_audio/output.wav 0 → 100644
View file @ e3cdb305
File suppressed by a .gitattributes entry or the file's encoding is unsupported.
requirements.txt 0 → 100644
View file @ e3cdb305
soundfile
librosa
sox
transformers==4.57.3
qwen-tts
\ No newline at end of file
run_server.sh 0 → 100644
View file @ e3cdb305
#!/usr/bin/env bash
set -euo pipefail
SERVER="${SERVER:-http://127.0.0.1:8000}"
MODEL="${MODEL:-qwen3-tts}"
LANGUAGE="${LANGUAGE:-Auto}"
curl -sS "${SERVER}/v1/audio/speech" \
-H "Content-Type: application/json" \
-o output.wav \
-d @- <<EOF
{
"model": "${MODEL}",
"text": "哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"task_type": "VoiceDesign",
"language": "${LANGUAGE}",
"instruct": "体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
"generation_params": {
"max_new_tokens": 4096,
"do_sample": true,
"top_k": 50,
"top_p": 1.0,
"temperature": 0.9
},
"response_format": "wav"
}
EOF
test_model_12hz_base.py 0 → 100644
View file @ e3cdb305
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def ensure_dir(d: str):
os.makedirs(d, exist_ok=True)
def run_case(tts: Qwen3TTSModel, out_dir: str, case_name: str, call_fn):
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = call_fn()
torch.cuda.synchronize()
t1 = time.time()
print(f"[{case_name}] time: {t1 - t0:.3f}s, n_wavs={len(wavs)}, sr={sr}")
for i, w in enumerate(wavs):
sf.write(os.path.join(out_dir, f"{case_name}_{i}.wav"), w, sr)
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-Base/"
OUT_DIR = "qwen3_tts_test_voice_clone_output_wav"
ensure_dir(OUT_DIR)
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# Reference audio(s)
ref_audio_path_1 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/clone_2.wav"
ref_audio_path_2 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/clone_1.wav"
ref_audio_single = ref_audio_path_1
ref_audio_batch = [ref_audio_path_1, ref_audio_path_2]
ref_text_single = "Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And thanks to you."
ref_text_batch = [
"Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And thanks to you.",
"甚至出现交易几乎停滞的情况。",
]
# Synthesis targets
syn_text_single = "Good one. Okay, fine, I'm just gonna leave this sock monkey here. Goodbye."
syn_lang_single = "Auto"
syn_text_batch = [
"Good one. Okay, fine, I'm just gonna leave this sock monkey here. Goodbye.",
"其实我真的有发现，我是一个特别善于观察别人情绪的人。",
]
syn_lang_batch = ["Chinese", "English"]
common_gen_kwargs = dict(
max_new_tokens=2048,
do_sample=True,
top_k=50,
top_p=1.0,
temperature=0.9,
repetition_penalty=1.05,
subtalker_dosample=True,
subtalker_top_k=50,
subtalker_top_p=1.0,
subtalker_temperature=0.9,
)
for xvec_only in [False, True]:
mode_tag = "xvec_only" if xvec_only else "icl"
# Case 1: prompt single + synth single, direct
run_case(
tts, OUT_DIR, f"case1_promptSingle_synSingle_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_single,
language=syn_lang_single,
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
**common_gen_kwargs,
),
)
# Case 1b: prompt single + synth single, via create_voice_clone_prompt
def _case1b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
)
return tts.generate_voice_clone(
text=syn_text_single,
language=syn_lang_single,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case1_promptSingle_synSingle_promptThenGen_{mode_tag}",
_case1b,
)
# Case 2: prompt single + synth batch, direct
run_case(
tts, OUT_DIR, f"case2_promptSingle_synBatch_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
**common_gen_kwargs,
),
)
# Case 2b: prompt single + synth batch, via create_voice_clone_prompt
def _case2b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
)
return tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case2_promptSingle_synBatch_promptThenGen_{mode_tag}",
_case2b,
)
# Case 3: prompt batch + synth batch, direct
run_case(
tts, OUT_DIR, f"case3_promptBatch_synBatch_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
ref_audio=ref_audio_batch,
ref_text=ref_text_batch,
x_vector_only_mode=[xvec_only, xvec_only],
**common_gen_kwargs,
),
)
# Case 3b: prompt batch + synth batch, via create_voice_clone_prompt
def _case3b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_batch,
ref_text=ref_text_batch,
x_vector_only_mode=[xvec_only, xvec_only],
)
return tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case3_promptBatch_synBatch_promptThenGen_{mode_tag}",
_case3b,
)
if __name__ == "__main__":
main()
test_model_12hz_custom_voice.py 0 → 100644
View file @ e3cdb305
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice/"
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# -------- Single (with instruct) --------
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_custom_voice(
text="其实我真的有发现，我是一个特别善于观察别人情绪的人。",
language="Chinese",
speaker="Vivian",
instruct="用特别愤怒的语气说",
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[CustomVoice Single] time: {t1 - t0:.3f}s")
sf.write("qwen3_tts_test_custom_single.wav", wavs[0], sr)
# -------- Batch (some empty instruct) --------
texts = ["其实我真的有发现，我是一个特别善于观察别人情绪的人。", "She said she would be here by noon."]
languages = ["Chinese", "English"]
speakers = ["Vivian", "Ryan"]
instructs = ["", "Very happy."]
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_custom_voice(
text=texts,
language=languages,
speaker=speakers,
instruct=instructs,
max_new_tokens=2048,
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[CustomVoice Batch] time: {t1 - t0:.3f}s")
for i, w in enumerate(wavs):
sf.write(f"qwen3_tts_test_custom_batch_{i}.wav", w, sr)
if __name__ == "__main__":
main()
test_model_12hz_voice_design.py 0 → 100644
View file @ e3cdb305
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign/"
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# -------- Single --------
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_voice_design(
text="哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
language="Chinese",
instruct="体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[VoiceDesign Single] time: {t1 - t0:.3f}s")
sf.write("qwen3_tts_test_voice_design_single.wav", wavs[0], sr)
# -------- Batch --------
texts = [
"哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"It's in the top drawer... wait, it's empty? No way, that's impossible! I'm sure I put it there!"
]
languages = ["Chinese", "English"]
instructs = [
"体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
"Speak in an incredulous tone, but with a hint of panic beginning to creep into your voice."
]
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_voice_design(
text=texts,
language=languages,
instruct=instructs,
max_new_tokens=2048,
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[VoiceDesign Batch] time: {t1 - t0:.3f}s")
for i, w in enumerate(wavs):
sf.write(f"qwen3_tts_test_voice_design_batch_{i}.wav", w, sr)
if __name__ == "__main__":
main()
\ No newline at end of file
torchaudio-2.5.1+das.opt1.dtk25042.20251127.g10a9ffcd-cp310-cp310-manylinux_2_28_x86_64.whl 0 → 100644
View file @ e3cdb305
File suppressed by a .gitattributes entry or the file's encoding is unsupported.
vllm.zip 0 → 100644
View file @ e3cdb305
File suppressed by a .gitattributes entry or the file's encoding is unsupported.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment