Show HN: PocketTTS-raven – fast local in-browser TTS with voice cloning

Category: ai-ml

Tags: text-to-speech, voice-cloning, wasm, onnx-runtime, real-time, performance-optimization

Score: 7.3/10 (Innovation: 6, Technical: 9, Documentation: 8, Utility: 7)

PocketTTS-raven is a highly optimized fork of PocketTTS that achieves faster-than-realtime text-to-speech with voice cloning, running locally in the browser via WASM and natively on CPU. It delivers impressive latency improvements (30ms first useful audio, 33x realtime native) through graph rewrites, cache deltas, and custom ops, making it ideal for low-latency conversational AI and interactive applications.

Target audience: AI developers, game developers, and researchers building real-time voice-enabled applications

Repository: https://pantel.is/projects/pocket-tts-raven/

View on Hacker News