Show HN: Recreate Thinking Machines 276B voice demo with duct tape and 8B model

Category: ai-ml

Tags: voice-agent, real-time, event-loop, vision, speech-recognition, llm

Score: 7.5/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 7)

Cheap-IM is a CPU-only voice agent that replicates the surface behaviors of Thinking Machines' advanced voice demo using off-the-shelf commodity models and a Python event loop. It cleverly combines local speech and vision processing with remote LLM calls to achieve real-time interaction, live translation, and proactive behaviors on a laptop. The project is interesting for its pragmatic approach to gluing together existing technologies to mimic cutting-edge demos with minimal resources.

Target audience: backend devs, ai engineers, devops

Repository: https://github.com/kouhxp/cheap-im · Python · 17 stars

View on Hacker News