inclusionAI/Ming-3.0-tts

选用指南

AI 根据 README 整理 · 2026/10/21 17:21:17

使用方式

提供命令行工具和 Python API 两种使用方式，支持多种任务模式，如 TTS、语音克隆、语音设计、多语言、方言、情感、属性控制、自由形式控制和播客等。

查看 README 依据

### <a id="ming-30-tts-05b"></a>Ming-3.0-tts-0.5b ```bash python infer.py \ --model_path inclusionAI/Ming-3.0-tts-0.5b \ --task TTS \ --language_id Chinese \ --text "傍晚的风从山口吹过来，远处的云慢慢染成了金色。" \ --prompt_wav_path data/voice_presets/voice_audio/lingguang.wav \ --prompt_text "诶，你今天有空吗？我们能不能聊聊这件事啊？" \ --use_spk_emb \ --output output/tts.wav ``` Or from Python: ```python from ming_tts import MingTTS model = MingTTS("inclusionAI/Ming-3.0-tts-0.5b") model.synthesize( task="TTS", language_id="Chinese", # English for English text

| `freeform` | none by default (recommended); optional | not accepted | — | `instruction` | | `voice_design` | none | — | — | `style` | | `ip` | none | — | — | `ip` | `python infer_tiny.py --help` lists the same arguments for the command line; repeat `--prompt_wav` and `--prompt_text` for podcast. | | Ming-3.0-tts-0.5b | Ming-3.0-tts-tiny | | --- | --- | --- | | Python API | `ming_tts.MingTTS` | `ming_tiny.MingTinyTTS` | | CLI | `infer.py` | `infer_tiny.py`, `examples/one_click/` | | Preflight | `tools/preflight.py` | `tools/preflight_tiny.py` |

部署要求

0.5b 模型环境已在 NVIDIA H800-80GB 和 H20-96G 上测试，flash-attn 可选；tiny 模型推荐使用 NVIDIA H800，需要 Python 3.10.15、PyTorch 和 torchaudio 2.7.1+cu118 以及 transformers==4.46.3。

查看 README 依据

├── tokenizer_config.json # special_tokens_map.json / chat_template.jinja └── README.md ``` ## <a id="quick-start-"></a>Quick Start 🚀 ### Installation ```bash git clone https://github.com/inclusionAI/Ming-3.0-tts.git cd Ming-3.0-tts pip install -r requirements.txt # Ming-3.0-tts-0.5b ``` Or build the 0.5b Docker image (torch 2.6, CUDA 12.4, `transformers==4.46.3`): ```bash docker build -f docker/Dockerfile -t ming-tts . ``` The 0.5b environment has been tested on NVIDIA H800-80GB and H20-96G. For 0.5b, `flash-attn` is optional; inference falls back to PyTorch SDPA.

Tiny uses a separate reference inference environment: Python **3.10.15**, PyTorch and torchaudio **2.7.1+cu118** (CUDA 11.8 wheels), and `transformers==4.46.3`. Install the pinned dependencies from [requirements-tiny.txt](requirements-tiny.txt) in a dedicated environment, as other dependency versions differ from 0.5b. We recommend using NVIDIA H800 for inference. ```bash conda create -n ming-tiny python=3.10.15 -y && conda activate ming-tiny pip install -r requirements-tiny.txt ``` ### Model Download ```bash pip install "huggingface_hub[cli]"

已知限制

能力使用单独的任务模式，不支持任意组合；freeform 任务不推荐添加参考音频，且不与语音设计、内置语音、情感、方言或属性控制组合使用。

查看 README 依据

- **Multi-speaker dialogue:** generate a conversation from a script and speaker references (**tiny only**). Capabilities use separate task modes, not arbitrary combinations. See the [Cookbook](#cookbook) for supported inputs, [Performance](#performance) for benchmarks, and [Blog & Demos](https://nomorecoding.github.io/Ming-3.0-tts/) for audio examples. ## <a id="demo-video"></a>Demo Video https://github.com/user-attachments/assets/4062ba4c-7d2d-4a45-a19b-760100b79f9c ## Architecture

text="The garden will remain open until sunset, provided the weather stays clear.", controls={"ip": "US_James"}, use_zero_spk_emb=True, ) ``` ### How it is said **Freeform**: describe the voice and delivery with one natural-language instruction. Both models default to generation without reference audio; adding a reference is not recommended. Not combined with voice design, a built-in voice, emotion, dialect or attribute controls. ```python model.synthesize( task="FREEFORM", text="The old radio crackled softly as the storm moved across the valley.",