Show HN: Varkos – a local AI pet that can talk, interact, and play small games

Category: ai-ml

Tags: text-to-speech, voice-cloning, onnx-runtime, wasm, cpp

Score: 7.0/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 6)

PocketTTS-RAVEN is a highly optimized C++ runtime for Kyutai's Pocket TTS model that enables faster-than-realtime voice cloning and text-to-speech on CPU without Python or GPU dependencies. It leverages offline ONNX graph rewrites and a custom driver around stock ONNX Runtime to achieve impressive performance in native and browser (WASM) environments, with an OpenAI-compatible API and C FFI.

Target audience: backend developers, AI engineers, embedded developers

Repository: https://varkos.party/ · C++ · MIT · 26 stars

View on Hacker News