[Tether AI Research Releases Edge-Side Speech Recognition Model TranscribePsy for Smart Glasses and Humanoid Robots] BlockBeats News, October 14 — Tether AI Research has released the open-source multilingual automatic speech recognition model QVAC TranscribePsy, with approximately 254 million parameters. It can run entirely on-device without requiring audio to be uploaded to the cloud. According to the official announcement, it can process audio from smart glasses connected via Bluetooth or Wi-Fi in near real-time and can also operate on onboard processors in humanoid robots. On the Core ML version for the iPhone 17 Pro Max, transcribing 10 minutes of audio takes approximately 1.65 seconds. Based on disclosed evaluation data, TranscribePsy achieved a Word Error Rate (WER) of 8.02% on the Open ASR Leaderboard, 6.58% on Multilingual LibriSpeech, and 21.56% on Common Voice. It is the fastest among similar models on six tested smartphones, approximately 2.6 times faster than the second-fastest model. The model supports 25 European languages on laptops and 8 European languages on mobile devices. It is released under the Apache 2.0 license, along with model weights and builds for GGUF, Core ML, and other formats. Tether CEO Paolo Ardoino stated that they hope advanced speech AI can run locally on users' own devices, ensuring data never leaves the device. [Original Link]