Open Source JavaScript Text-to-Speech (TTS) Models

Kokoro is an open-weight text-to-speech model and inference library built around the lightweight Kokoro-82M model. It is designed to generate high-quality speech from text while staying fast, compact, and cost-efficient compared with larger TTS systems. The project is useful for developers who want deployable speech synthesis without depending on a closed platform. It can be installed as a Python package and used in applications, scripts, experiments, or production workflows. Kokoro’s Apache-licensed weights make it flexible for personal, research, and commercial use cases. Its main value is delivering practical, accessible voice generation with a small model footprint and a simple developer workflow.

Supertonic is a lightning-fast, on-device text-to-speech system built around ONNX Runtime for maximum speed and portability. It focuses on running entirely locally, eliminating the need for cloud APIs and providing low latency and strong privacy guarantees, even on constrained devices like Raspberry Pi boards and e-readers. The core model is highly compact at around 66 million parameters, yet benchmarks show it can generate speech up to 167× faster than real time on modern consumer hardware and significantly outpace popular cloud TTS APIs in throughput and real-time factor. Supertonic is designed to handle real-world text gracefully, including numbers, dates, currency symbols, abbreviations, and technical units, without requiring heavy pre-processing or custom text normalization. The repository provides complete reference implementations across many programming ecosystems—Python, Node.js, browser (WebGPU/WASM), Java, C++, C#, Go, Swift, iOS, Rust, and Flutter.

Text-to-speech software developed by Electron + vue + ElementUI + js. The high-fidelity and flexible configuration of speech synthesis products opens up the closed loop of human-computer interaction and enables applications to sound realistically. A variety of timbres are available, and functions such as adjusting speech rate, intonation, and volume are provided. Technically, multi-level rhythmic pauses are taken into account to achieve the purpose of natural synthesizing rhythm, and comprehensively use acoustic parameters and linguistic parameters to establish multiple automatic prediction models based on deep learning. Using massive audio data to train the pronunciation model, the synthetic sound is real, full, cadenced, and expressive, and the MOS score has reached the professional level in the industry.