Skip to content

Releases: wildminder/ComfyUI-KaniTTS

Release list

v1.0.0 - Initial Release

v1.0.0 - Initial Release Latest
Latest

Choose a tag to compare

View all tags
@wildminder wildminder released this 02 Oct 18:20
1.0.0
5e3de00
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

This marks the initial public release of ComfyUI-KaniTTS, a custom node for integrating the KaniTTS family of text-to-speech models into ComfyUI workflows.

This release provides a stable foundation for generating high-quality speech with robust model management and control over generation parameters.

Key Features

  • Multi-Model Support: Full integration of five different KaniTTS models, including the multi-speaker 370m variant and several 450m base and fine-tuned models.
  • Speaker Selection: For the kani-tts-370m model, users can select from a list of 15 predefined voices across multiple languages.
  • Automatic Model Management: The node handles the automatic download and caching of all required models, including KaniTTS transformers and the NeMo audio codec dependency, into the ComfyUI models/tts directory.
  • VRAM Efficiency: All models are managed by ComfyUI's core systems, ensuring they are loaded to the GPU for inference and offloaded correctly to conserve VRAM.
  • Generation Control: Provides UI controls for key generation parameters such as temperature, top-p, repetition penalty, and seed for reproducible outputs.

Supported Models

This version supports the following KaniTTS models:

Model Name Parameters Description / Key Features
kani-tts-370m 370M Multi-Speaker. Supports a wide range of predefined voices.
kani-tts-450m-0.1-pt 450M Base Model. Generates a generic/randomized voice.
kani-tts-450m-0.1-ft 450M Finetuned (Male). Produces a consistent male voice.
kani-tts-450m-0.2-pt 450M Base Model 2. Broader multilingual support.
kani-tts-450m-0.2-ft 450M Finetuned (Female). Produces a consistent female voice.

🔧 Installation & Setup

via the ComfyUI Manager or do it manually:

Windows (Manual Installation Recommended)

The automatic installation of nemo_toolkit often fails on Windows. It is recommended to manually install the pre-built packages (.whl files) for your Python version.

  1. Download Wheels: Obtain the required .whl files for your Python version from the Python-Windows-WHL Hugging Face repository.

    Package Name Version Python Version
    nemo_toolkit 2.6.0rc0 3.12 / 3.13
    pynini 2.1.6 / 2.1.7 3.12 / 3.13
    editdistance 0.8.1 3.13
    megatron_core 0.13.1 3.12 / 3.13
    texterrors 1.0.9 3.12 / 3.13

  2. Install Wheels: Open a command prompt, navigate to your downloads folder, and run pip install on each file. See the project README.md for the exact commands for your Python version.

Assets 2
Loading