Top AI Repos — open-source AI, indexed and scored

Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.

HuiResearch/FlashTTS

基于SparkTTS、OrpheusTTS等模型，提供高质量中文语音合成与声音克隆服务。

20No licencePython

View on GitHub

Stars: 612
Forks: 76
Watchers: 612
Open issues: 47
Contributors: 2

Star history, last 90 days— over 5 snapshots

2026-07-24 · 6122026-08-06 · 612

Star history, last 90 days
Date	Stars
2026-07-24	612
2026-07-25	612
2026-07-28	612
2026-07-30	612
2026-08-06	612

Today

— stars today

This week

— stars this week

This month

— stars this month

Momentum

0.0

growth rate 0.00%/day

From the README

<div align="center">
  <img src="docs/zh/_img/icon.png" width="450" alt="FlashTTS Logo"/>

[📘 Documentation](docs/zh/README.MD) | [📚 Deepwiki](https://deepwiki.com/HuiResearch/FlashTTS)

[中文](README.MD) | [English](README_EN.MD)

</div>

---

# FlashTTS

> 基于 **SparkTTS、OrpheusTTS、MegaTTS&nbsp;3** 等先进模型，FlashTTS 提供高质量的中文语音合成与零样本声音克隆服务。借助简洁直观的
> Web 界面，您可以快速生成自然逼真的人声，满足配音、阅读、无障碍、虚拟角色等多元场景需求。

> 如果 FlashTTS 对你有帮助，欢迎留下一个 ⭐ Star！

## ✨ 功能亮点

|     | 功能          | 说明                                                                |
|-----|-------------|-------------------------------------------------------------------|
| 🚀  | **多后端推理加速** | 支持 `vllm`、`sglang`、`llama‑cpp`、`mlx‑lm`、`tensorrt-llm` 等多种高性能推理引擎 |
| 🎯  | **高并发**     | 动态批处理与异步队列，轻松应对大流量请求                                              |
| 🎛️ | **全参数控制**   | 可调节音调、语速、温度、情感标签等                                                 |
| 📱  | **轻量部署**    | 基于 FastAPI，一条命令即可启动；最小依赖                                          |
| 🔊  | **长文本合成**   | 支持超长文本，保持连续音色一致                                                   |
| 🔄  | **流式 TTS**  | 边生成边播放，降低等待，提高交互体验                                                |
| 🎭  | **多角色对话**   | 同文本多角色合成，适合剧本配音                                                   |
| 🎨  | **现代化前端**   | 适配Web端                                                            |

## 🖼️ 前端示例

https://github.com/user-attachments/assets/1bd9d586-fac7-4016-b955-5a58d8fb9d7e

## 🔈 语音演示

下列 Demo 展示了 FlashTTS 在不同模型与角色上的克隆效果。

### SparkTTS 模型

<table>
<tr>
<td align="center">

**Donald Trump en**
</td>
<td align="center">

**Donald Trump zh**
</td>
</tr>

<tr>
<td align="center">

[Donald Trump en](https://github.com/user-attachments/assets/79d4d1d1-5579-4ddb-8092-500b3db05d45)

</td>
<td align="center">

[Donald Trump zh](https://github.com/user-attachments/assets/262f4306-4a0f-4735-b4e7-708cd3193324)

</td>
</tr>
</table>

<table>
<tr>
<td align="center">

**哪吒**
</td>
<td align="center">

**李靖**
</td>
</tr>

<tr>
<td align="center">

[哪吒](https://github.com/user-attachments/assets/4743778b-63c7-4c42-95b2-636828364d67)

</td>
<td align="center">

[李靖](https://github.com/user-attachments/assets/b5a51fa9-c5a1-4f39-bbf8-09e73b2c1936)

</td>
</tr>
</table>

<table>
<tr>
<td align="center">

**余承东**
</td>
<td align="center">

**徐志胜**
</td>
</tr>

<tr>
<td align="center">

[余承东](https://github.com/user-attachments/assets/1cd5f156-010d-4bc9-8e99-f37def54b99f)

</td>
<td align="center">

[徐志胜](https://github.com/user-attachments/assets/f197ac39-5928-4ead-83d7-071b1bf5f974)

</td>
</tr>
</table>

### MegaTTS&nbsp;3 模型

<table>
<tr>
<td align="center">

**蔡徐坤**
</td>
<td align="center">

**太乙真人**
</td>
</tr>

<tr>
<td align="center">

[蔡徐坤](https://github.com/user-attachments/assets/bcb48ead-f157-45d7-8060-cbccf418f0ed)

</td>
<td align="center">

[太乙真人](https://github.com/user-attachments/assets/73edbd6b-ece3-4e22-95a2-53b7c4e45346)

</td>
</tr>
</table>

### OrpheusTTS (zh) 模型

<table>
<tr>
<td align="center">

**长乐**
</td>
<td align="center">

**白芷**
</td>
</tr>

<tr>
<td align="center">

[长乐](https://github.com/user-attachments/assets/271bd5a3-dd56-430a-87d0-ebea0e397a89)

</td>
<td align="center">

[白芷](https://github.com/user-attachments/assets/e012c580-840f-401c-99f5-665c72f6b89f)

</td>
</tr>
</table>

## 快速开始

建议在 **python 3.8 - 3.12** 环境中使用 pip 安装 `flashtts`：

```bash
pip install flashtts
```

详细安装步骤，请参考：[installation.md](docs/zh/get_started/installation.md)

本地推理：
```bash
flashtts infer \
  -i "你好，欢迎使用语音合成。" \
  -o output.wav \
  -m ./models/your_model \
  -b vllm \
  [其他可选参数]
```

详细使用说明，请参考：[quick_start.md](docs/zh/get_started/quick_start.md)

部署方式：
```bash
 flashtts serve \
 --model_path Spark-TTS-0.5B \ 
 --backend vllm \ 
 --role_dir data/roles \ 
 --llm_device cuda \
 --tokenizer_device cuda \
 --detokenizer_device cuda \
 --wav2vec_attn_implementation sdpa \
 --llm_attn_implementation sdpa \ 
 --torch_dtype "bfloat16" \ 
 --max_length 32768 \
 --llm_gpu_memory_utilization 0.6 \
 --fix_voice

Excerpt of 5,808 characters

Read on GitHub

Top contributors2 shown

HuiResearch
124
dylanoy
1

Quality score

Would you bet a product on this? Bounded 0–100 and slow moving.

20/10020

Maintenance: 5%; How recently the default branch was pushed to
Releases: 10%; Recency and cadence of tagged releases
Community: 9%; Contributor breadth and bus factor
Issue backlog: 35%; Open issues measured against the size of the audience
Documentation: 76%; README depth, plus a homepage and topics
Licensing: 0%; How freely the code can be adopted

What to know before adopting

No commits in over a year
No release in over a year
Bus factor of 1 — one contributor dominates
No license — all rights reserved by default, not safe to reuse
No licence file — default copyright applies, all rights reserved.

Details

Created: last year
Last push: last year
Last release: last year
Release tag: v0.1.7
Releases in the last year: 0
Default branch: master
Size: 33k KB
Owner type: User
Owner location: —
Discovered via: topic:vllm
First indexed: 14 days ago
Last synced: 12 days ago

HuiResearch/FlashTTS

基于SparkTTS、OrpheusTTS等模型，提供高质量中文语音合成与声音克隆服务。

20No licencePython

View on GitHub

Stars: 612
Forks: 76
Watchers: 612
Open issues: 47
Contributors: 2

Star history, last 90 days— over 5 snapshots

2026-07-24 · 6122026-08-06 · 612

Star history, last 90 days
Date	Stars
2026-07-24	612
2026-07-25	612
2026-07-28	612
2026-07-30	612
2026-08-06	612

Today

— stars today

This week

— stars this week

This month

— stars this month

Momentum

0.0

growth rate 0.00%/day

From the README

<div align="center">
  <img src="docs/zh/_img/icon.png" width="450" alt="FlashTTS Logo"/>

[📘 Documentation](docs/zh/README.MD) | [📚 Deepwiki](https://deepwiki.com/HuiResearch/FlashTTS)

[中文](README.MD) | [English](README_EN.MD)

</div>

---

# FlashTTS

> 基于 **SparkTTS、OrpheusTTS、MegaTTS&nbsp;3** 等先进模型，FlashTTS 提供高质量的中文语音合成与零样本声音克隆服务。借助简洁直观的
> Web 界面，您可以快速生成自然逼真的人声，满足配音、阅读、无障碍、虚拟角色等多元场景需求。

> 如果 FlashTTS 对你有帮助，欢迎留下一个 ⭐ Star！

## ✨ 功能亮点

|     | 功能          | 说明                                                                |
|-----|-------------|-------------------------------------------------------------------|
| 🚀  | **多后端推理加速** | 支持 `vllm`、`sglang`、`llama‑cpp`、`mlx‑lm`、`tensorrt-llm` 等多种高性能推理引擎 |
| 🎯  | **高并发**     | 动态批处理与异步队列，轻松应对大流量请求                                              |
| 🎛️ | **全参数控制**   | 可调节音调、语速、温度、情感标签等                                                 |
| 📱  | **轻量部署**    | 基于 FastAPI，一条命令即可启动；最小依赖                                          |
| 🔊  | **长文本合成**   | 支持超长文本，保持连续音色一致                                                   |
| 🔄  | **流式 TTS**  | 边生成边播放，降低等待，提高交互体验                                                |
| 🎭  | **多角色对话**   | 同文本多角色合成，适合剧本配音                                                   |
| 🎨  | **现代化前端**   | 适配Web端                                                            |

## 🖼️ 前端示例

https://github.com/user-attachments/assets/1bd9d586-fac7-4016-b955-5a58d8fb9d7e

## 🔈 语音演示

下列 Demo 展示了 FlashTTS 在不同模型与角色上的克隆效果。

### SparkTTS 模型

<table>
<tr>
<td align="center">

**Donald Trump en**
</td>
<td align="center">

**Donald Trump zh**
</td>
</tr>

<tr>
<td align="center">

[Donald Trump en](https://github.com/user-attachments/assets/79d4d1d1-5579-4ddb-8092-500b3db05d45)

</td>
<td align="center">

[Donald Trump zh](https://github.com/user-attachments/assets/262f4306-4a0f-4735-b4e7-708cd3193324)

</td>
</tr>
</table>

<table>
<tr>
<td align="center">

**哪吒**
</td>
<td align="center">

**李靖**
</td>
</tr>

<tr>
<td align="center">

[哪吒](https://github.com/user-attachments/assets/4743778b-63c7-4c42-95b2-636828364d67)

</td>
<td align="center">

[李靖](https://github.com/user-attachments/assets/b5a51fa9-c5a1-4f39-bbf8-09e73b2c1936)

</td>
</tr>
</table>

<table>
<tr>
<td align="center">

**余承东**
</td>
<td align="center">

**徐志胜**
</td>
</tr>

<tr>
<td align="center">

[余承东](https://github.com/user-attachments/assets/1cd5f156-010d-4bc9-8e99-f37def54b99f)

</td>
<td align="center">

[徐志胜](https://github.com/user-attachments/assets/f197ac39-5928-4ead-83d7-071b1bf5f974)

</td>
</tr>
</table>

### MegaTTS&nbsp;3 模型

<table>
<tr>
<td align="center">

**蔡徐坤**
</td>
<td align="center">

**太乙真人**
</td>
</tr>

<tr>
<td align="center">

[蔡徐坤](https://github.com/user-attachments/assets/bcb48ead-f157-45d7-8060-cbccf418f0ed)

</td>
<td align="center">

[太乙真人](https://github.com/user-attachments/assets/73edbd6b-ece3-4e22-95a2-53b7c4e45346)

</td>
</tr>
</table>

### OrpheusTTS (zh) 模型

<table>
<tr>
<td align="center">

**长乐**
</td>
<td align="center">

**白芷**
</td>
</tr>

<tr>
<td align="center">

[长乐](https://github.com/user-attachments/assets/271bd5a3-dd56-430a-87d0-ebea0e397a89)

</td>
<td align="center">

[白芷](https://github.com/user-attachments/assets/e012c580-840f-401c-99f5-665c72f6b89f)

</td>
</tr>
</table>

## 快速开始

建议在 **python 3.8 - 3.12** 环境中使用 pip 安装 `flashtts`：

```bash
pip install flashtts
```

详细安装步骤，请参考：[installation.md](docs/zh/get_started/installation.md)

本地推理：
```bash
flashtts infer \
  -i "你好，欢迎使用语音合成。" \
  -o output.wav \
  -m ./models/your_model \
  -b vllm \
  [其他可选参数]
```

详细使用说明，请参考：[quick_start.md](docs/zh/get_started/quick_start.md)

部署方式：
```bash
 flashtts serve \
 --model_path Spark-TTS-0.5B \ 
 --backend vllm \ 
 --role_dir data/roles \ 
 --llm_device cuda \
 --tokenizer_device cuda \
 --detokenizer_device cuda \
 --wav2vec_attn_implementation sdpa \
 --llm_attn_implementation sdpa \ 
 --torch_dtype "bfloat16" \ 
 --max_length 32768 \
 --llm_gpu_memory_utilization 0.6 \
 --fix_voice

Excerpt of 5,808 characters

Read on GitHub

Top contributors2 shown

HuiResearch
124
dylanoy
1

Quality score

Would you bet a product on this? Bounded 0–100 and slow moving.

20/10020

Maintenance: 5%; How recently the default branch was pushed to
Releases: 10%; Recency and cadence of tagged releases
Community: 9%; Contributor breadth and bus factor
Issue backlog: 35%; Open issues measured against the size of the audience
Documentation: 76%; README depth, plus a homepage and topics
Licensing: 0%; How freely the code can be adopted

What to know before adopting

No commits in over a year
No release in over a year
Bus factor of 1 — one contributor dominates

Details

Created: last year
Last push: last year
Last release: last year
Release tag: v0.1.7
Releases in the last year: 0
Default branch: master
Size: 33k KB
Owner type: User
Owner location: —
Discovered via: topic:vllm
First indexed: 14 days ago
Last synced: 12 days ago

Top AI Repos — open-source AI, indexed and scored

HuiResearch/FlashTTS

From the README

Top contributors2 shown

Quality score

What to know before adopting

Categories

Details

HuiResearch/FlashTTS

From the README

Top contributors2 shown

Quality score

What to know before adopting

Categories

Details

Top AI Repos — open-source AI, indexed and scored

Trend

From the README

Top contributors2 shown

Quality score

What to know before adopting

Categories

Details

Trend

From the README

Top contributors2 shown

Quality score

What to know before adopting

Categories

Details