Tether AI releases ultra-fast speech-to-text model TranscribePsy

On October 14, Tether CEO Paolo Ardoino revealed on Twitter that the QVAC platform under Tether AI has officially released the speech-to-text model TranscribePsy, featuring 250 million parameters and support for 25 European languages, specifically optimized for local execution on smartphones and laptops. On the iPhone 17 Pro Max neural engine, a 10-minute recording takes only 1.65 seconds to transcribe. The model requires no cloud support, adopts a decentralized peer-to-peer architecture, and emphasizes local processing, high privacy, and cross-hardware compatibility.