Show HN: Giving my inner monologue autocomplete, with my own voice

Category: ai-ml

Tags: voice-cloning, local-first, llm, speech-to-text, text-to-speech, browser-based

Score: 6.0/10 (Innovation: 7, Technical: 7, Documentation: 3, Utility: 4)

A browser-based application that captures your speech via VAD and STT, feeds it to a fine-tuned LLM (LFM2.5-350 trained on Simone Weil and Albert Camus notebooks), and generates possible continuations of your thoughts, spoken back in your cloned voice using Pocket TTS. The entire pipeline runs locally, enabling a real-time, uncanny experience of hearing your inner monologue with autocomplete.

Target audience: AI enthusiasts, researchers, and developers interested in local voice interfaces and creative LLM applications

Repository: https://www.maximevidal.com/voices

View on Hacker News