Show HN: Cheap-IM – CPU-only voice agent approximating Thinking Machines' demo

Category: ai-ml

Tags: voice-agent, real-time, cpu-only, asyncio, vision, speech-recognition

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

Cheap-IM is a CPU-only voice agent that replicates the interactive behaviors of a recent research demo by gluing commodity AI models (local ASR, vision, TTS plus cloud LLMs) together with a Python event loop. It is interesting for proving that a single laptop can achieve real-time speech, vision-triggered proactivity, live translation, and background tasks, and for its thoughtful architectural decisions around latency and minimal LLM calls.

Target audience: ai-ml engineers, backend developers, voice interface developers

Repository: https://github.com/kouhxp/cheap-im · Python

View on Hacker News