IndexTTS-2 - Expressive TTS & Voice Cloning - One Click Installer

IndexTTS-2 is our next-generation text-to-speech and voice cloning toolkit that makes it easy to turn text into expressive, natural-sounding speech. With two one-click Windows installers, you can set up Gradio or ComfyUI workflows quickly — no setup headaches, just one click and you're ready to generate.

IndexTTS-2 is an expressive TTS and voice cloning system built to deliver high-quality, emotion-aware speech. It supports voice cloning by using reference audio to tailor output to a desired voice.

Index TTS Original GitHub Repo: https://github.com/index-tts/index-tts

ComfyUI IndexTTS-2 Custom Node GitHub Repo: https://github.com/snicolast/ComfyUI-IndexTTS2

Gradio Installer Package — Automatic Setup

• Deepspeed 13.1
• Miniconda with pynini==2.1.6
• Triton for Windows
• PyTorch 2.8.0+cu128
• Default models downloaded and placed into their folders

ComfyUI Installer Package — Automatic Setup

• Sage Attention 2
• Flash Attention 2
• Miniconda with pynini==2.1.6
• Triton for Windows
• PyTorch 2.8.0+cu128
• Default models downloaded and placed into their folders

System Requirements

• GPU: Nvidia RTX 4090, 5090 series, or equivalent (for best performance)
• CUDA-compatible GPU with at least 8 GB VRAM
• Operating System: Windows
• Storage: At least 40 GB free
• FFmpeg: https://www.ffmpeg.org/download.html

Usage Notes

• Create a dedicated folder for the installer files and run the installer from that folder.
• Gradio: upload a reference audio, type the text to synthesize, then click the synthesize button.
• ComfyUI: upload a reference audio, enter text, and press the run button.
• Higher emotion strength can reduce similarity to the reference voice. Lower it if you want the output to more closely resemble the reference.
• Both interfaces allow you to upload a different reference audio for the desired emotion/style.
• To bypass reference audio in ComfyUI, right-click the reference nodes and select bypass.