echogarden-project/echogarden

Implements speech processing via pure TypeScript, WebAssembly, and the ONNX runtime—eliminating Python/Docker dependencies—with pluggable engine support for both offline models (Whisper, Kokoro, VITS) and cloud services (Google, Microsoft, OpenAI, AWS). Beyond basic synthesis and recognition, offers advanced capabilities like dynamic time warping alignment across 100+ languages, multi-language translation with word-level timing, voice isolation via source separation, and pronunciation enhancement through text normalization and lexicon support. An internal package manager auto-provisions voices and models on-demand, making it deployable as either a self-contained CLI tool or npm library targeting Node.js v18+.

439 stars and 7,824 monthly downloads. No commits in the last 6 months. Available on npm.