ComfyUI-Supertonic3TTS

custom nodes integrating Supertone's Supertonic-3 — a lightning-fast, on-device, multilingual Text-to-Speech system running natively via ONNX Runtime.

By Anonymzx·Created 3 months ago·Updated about a month ago· 2

Anonymzx/ComfyUI-Supertonic3TTS

Nodes—

On cloudLocal install

Stars2

Updatedabout a month ago

Readme

ComfyUI-Supertonic3TTS

A suite of ComfyUI custom nodes integrating Supertone's Supertonic-3 — a lightning-fast, on-device, multilingual Text-to-Speech system running natively via ONNX Runtime.

📦 Nodes

| Node | Input | Output | Description | |------|-------|--------|-------------| | Supertonic Model Loader 🎤 | — | SUPERTONIC_MODEL | Initialises the TTS engine. Auto-downloads ~400MB model on first run, stored locally in models/. | | Supertonic Text-to-Speech 🗣️ | model, text, lang, voice, speed, steps | AUDIO | Synthesises speech from text. 31 languages, 10 preset voices, expression tags, custom style .json via custom_style_path. | | Supertonic Effects ✨ | audio | AUDIO | Optional post-processing (trim, normalize, pitch, stretch, chorus). Apply to any AUDIO source. |

✨ Features

Supertonic TTS

31 languages — en, ko, ja, id, ar, de, es, fr, hi, vi, and more
10 built-in voices — M1–M5 (male), F1–F5 (female)
Custom voice styles — Pass an absolute path to a Supertonic .json voice profile in the custom_style_path field (e.g. from Supertone's Voice Builder)
Expression tags — Type tags like <laugh> or <sigh> directly into text for vocal expressions
Speed control — Native SDK speed parameter (0.5x – 2.0x). For finer post-synthesis tempo tweaks, use the SupertonicEffects time_stretch slider.
Steps — Diffusion steps (5–12, default 8). Higher = smoother, slower.
CPU-friendly — Runs entirely on CPU via ONNX Runtime, no GPU required
Local model cache — Model files stored in models/supertonic-3/ inside this node's directory (not ~/.cache/)

Pipeline

┌──────────────┐    ┌────────────────────────────┐    ┌──────────────┐
│   text +     │ →  │     SupertonicTTS          │ →  │   AUDIO      │
│  language    │    │ (lang, voice, speed, steps)│    │              │
└──────────────┘    └─────────────┬──────────────┘    └──────┬───────┘
                                  │                            │
                                  │       ┌──────────────┐     │
                                  └─────→ │ SupertonicEffects│ ←─┘
                                          │ (trim/pitch/etc)│
                                          └──────┬────────┘
                                                 ▼
                                          Preview / Save Audio

🎭 Expression Tags

Type any of the following tags directly into your text to add vocal expressions:

| Tag | Effect | |-----|--------| | <laugh> | Laughter | | <breath> | Breath intake | | <surprise> | Surprise tone | | <sigh> | Sigh | | <scream> | Scream / shout | | <throatclear> | Throat clear | | <sad> | Sad tone | | <angry> | Angry tone | | <cough> | Cough | | <yawn> | Yawn |

Example:

Halo! <laugh> Senang bertemu denganmu! <sigh> Tapi aku lelah.

The tags are passed to the Supertonic SDK which interprets them during synthesis.

🚀 Installation

Requirements

ComfyUI (any recent version)
Python 3.10+

1. Clone the repository

cd ComfyUI/custom_nodes/
git clone https://github.com/Anonymzx/ComfyUI-Supertonic3TTS.git

2. Install Python dependencies

Activate your ComfyUI virtual environment first, then:

pip install -r ComfyUI-Supertonic3TTS/requirements.txt

ComfyUI Manager is already!

you can install it on comfyui-manager!

Note: torch and torchaudio are not listed in requirements.txt — they are inherited from ComfyUI itself. Only supertonic, numpy, soundfile, and librosa are required on top of ComfyUI's base dependencies.

3. Restart ComfyUI

The nodes will appear under audio/Supertonic in the node menu.

First run only — the Loader auto-downloads the ~400MB Supertonic-3 model into models/supertonic-3/. A clean 0–100% slider shows progress in the console. Subsequent loads skip the download.

🎮 Usage

Basic TTS Pipeline

Add Supertonic Model Loader (no inputs needed)
Add Supertonic Text-to-Speech
Connect the model from step 1
Type your text (include expression tags like <laugh> if desired)
Configure: language, voice, speed, steps
Add Preview Audio or Save Audio to hear the result
Run the workflow

Add post-processing

Wire the TTS AUDIO output into Supertonic Effects, then into Preview/Save. Adjust pitch / stretch / chorus as needed.

Use a custom voice style

Set the optional custom_style_path field on the Supertonic Text-to-Speech node to an absolute path of a Supertonic voice style .json (e.g. generated from Supertone's Voice Builder). When set, it overrides the preset voice_style dropdown.

Speed vs Time Stretch

SupertonicTTS speed changes tempo during synthesis (model-aware, cleanest).
SupertonicEffects time_stretch uses phase vocoder after synthesis (any source, slight artifacts at extremes).
Combine both: SDK speed first, then Effects time_stretch on the output. Effective tempo ≈ speed × time_stretch.

🌍 Supported Languages

| Code | Language | Code | Language | Code | Language | |------|----------|------|----------|------|----------| | en | English | ko | Korean | ja | Japanese | | ar | Arabic | bg | Bulgarian | cs | Czech | | da | Danish | de | German | el | Greek | | es | Spanish | et | Estonian | fi | Finnish | | fr | French | hi | Hindi | hr | Croatian | | hu | Hungarian | id | Indonesian | it | Italian | | lt | Lithuanian | lv | Latvian | nl | Dutch | | pl | Polish | pt | Portuguese | ro | Romanian | | ru | Russian | sk | Slovak | sl | Slovenian | | sv | Swedish | tr | Turkish | uk | Ukrainian | | vi | Vietnamese | na | Unknown / fallback | |

📄 License

Code: MIT License Model: OpenRAIL-M License (Supertone)