index-tts /index-tts

About index-tts

IndexTTS copies a voice from one short clip, then reads your script in it. Version 2.5 covers Chinese, English, Japanese, Spanish and Arabic, and you can steer emotion, speed and hard-word pronunciation. An NVIDIA GPU with CUDA 12.8 or newer is required. The license is custom, so check it before commercial use.

Best for: Short-drama and comic-drama dubbers who need cloned voices in five languages.

• Clones a voice from one reference clip, with no training per speaker
• Sets emotion with a sad or happy reference clip, an 8-value vector, or text
• emo_alpha (0.0-1.0) controls how strongly the emotion reference pulls the voice
• Changes speaking speed with duration_factor, from 0.5x to 2.0x
• Fixes pronunciation with Pinyin, CMU phonemes or Japanese Kana
• Runs as a Gradio WebUI, a Python API, or a vLLM server

1. 1
   Clone the repo: the command below

   git clone https://github.com/index-tts/index-tts.git && cd index-tts

   Copy
2. 2
   Install uv, then sync dependencies: pip install -U uv and uv sync --all-extras
3. 3
   Download the weights: the command below

   hf download IndexTeam/IndexTTS-2.5 --local-dir=checkpoints

   Copy
4. 4
   Check your GPU with uv run tools/gpu_check.py
5. 5
   Start the WebUI with uv run webui.py, then open http://127.0.0.1:7860
6. 6
   Or run the CLI: the command below

   PYTHONPATH="$PYTHONPATH:." uv run indextts/infer_v2_5.py --cfg_path checkpoints/config.yaml --model_dir checkpoints --text "Hello world" --lang EN

   Copy

• index-tts vs GPT-SoVITS