@vocoloco/engine

WebGPU inference for OmniVoice text-to-speech, its audio codec and Whisper speech recognition, plus a Qwen2 BPE tokenizer. Plain ES modules, no dependencies. Model weights are not included.