Tether Brings On-Device Speech Recognition to Smart Glasses and Humanoid Robots as Smart Technology Moves Deeper Into Private Life

TranscribePsy’s open-source 254-million-parameter model delivers high-performance speech recognition without cloud processing, transcribing 10 minutes of audio in 1.65 seconds on an iPhone 17 Pro Max and advances Tether’s vision for private, voice-enabled AI

14 October 2026 – Tether AI Research* today announced the release of QVAC TranscribePsy, a compact 254-million-parameter multilingual automatic speech recognition model designed to run entirely on-device. The model marks a major step in Tether’s broader effort to bring advanced AI capabilities directly to smartphones, smart glasses and humanoid robots while keeping users’ data private.

As voice-enabled technology moves further into homes, workplaces and everyday conversations, TranscribePsy gives developers a way to build speech recognition that works offline without sending users’ audio to cloud servers. The model can process audio from smart glasses connected via Bluetooth or Wi-Fi in near real-time, or run directly on a humanoid robot’s onboard processor. Designed for fast responses, it enables natural voice interactions with low latency, while keeping speech processing on-device.

The release addresses a growing question for consumer AI: as more devices can listen, where do those conversations go? Spoken interactions can capture family discussions, confidential work, and the voices of people nearby who never chose to use the technology. Processing speech locally removes the need to upload that audio for transcription, giving people greater control over information from their private lives.

High Performance in a Compact Model

TranscribePsy delivers leading performance across accuracy, speed, and memory efficiency, despite being a fraction of the size of larger speech models.

• TranscribePsy achieves an 8.02% word error rate (WER) on the Open ASR Leaderboard, 6.58% on Multilingual LibriSpeech and 21.56% on Common Voice, leading all tested models in its parameter class.
• At 254M parameters, TranscribePsy achieves a 6.58% WER on Multilingual LibriSpeech, outperforming models 3–4x its size, including Qwen3-ASR-0.6B, omniASR-$CTC-1B-v2, OWSM-$CTC-v4 and Whisper-Medium.
• Across six tested smartphones, TranscribePsy was the fastest model, running 2.6x faster than the next-fastest model and 5.7x faster than Whisper-small across the phone CPUs.
• Its Core ML build transcribes 10 minutes of audio in 1.65 seconds on an iPhone 17 Pro Max, while using less than half the disk space and 1.7x less memory than a tested Core ML competitor.
• TranscribePsy covers 25 European languages on laptops and 8 European languages on phones, with dedicated builds for the iPhone Neural Engine.

Built for Voice-First AI

TranscribePsy is designed for continuous, real-time audio, turning conversations into text moments after they are spoken. These models support Tether’s longer-term vision for AI that people can interact with AI naturally through speech. For smart glasses, local transcription can reduce delays and data leaks associated with cloud processing. This approach is particularly relevant to medical consultations, legal interviews, therapy sessions and confidential journalism, where recordings may contain sensitive information. In humanoid robotics, a compact speech model can preserve onboard computing resources for visual understanding, spatial mapping and physical control.

The release includes four specialized checkpoints: a laptop model covering 25 European languages, a laptop variant focused on eight core languages, a GGUF mobile build that runs on the qvac-fabric-speech’s parakeet.cpp engine, and a Core ML build optimized for the iPhone Neural Engine.

Within the broader QVAC Psy ecosystem, TranscribePsy provides the speech-to-text foundation for potential integrations with VisionPsy, TranslatePsy-Nano and MedPsy. Together, these capabilities could support voice-driven document analysis, translation, navigation and private note-taking. They reflect Tether’s broader ambition to bring private, on-device AI into everyday tasks, with this release delivering the speech recognition models that lay the foundation for those future applications.

“The future of AI shouldn’t mean giving centralized platforms a front-row seat to our personal lives,” said Paolo Ardoino, CEO of Tether. “TranscribePsy shows that advanced speech AI can be fast, capable, and entirely local. AI is going to become part of our phones, glasses, and machines, but that doesn’t mean our data needs to leave them. We want to see powerful AI running directly on the devices people already own, with users in control of their data and their privacy intact.”

Open by Design

TranscribePsy is released under the Apache 2.0 license for research and educational purposes, with open model weights, GGUF mobile builds, Apple Silicon Core ML builds, and complete evaluation configurations for transparent and reproducible benchmarking.

The release is part of Tether AI Research’s broader work to develop compact, efficient models capable of processing audio, text, images, and video directly on edge devices. The long-term vision is to make advanced multimodal AI available wherever people need it, without requiring sensitive data to leave the device.

*References to Tether AI Research mean Tether Data, S.A. de C.V.

About QVAC

QVAC is Tether Data’s advanced AI research initiative dedicated to building open, decentralized, and adaptive intelligence systems. Its mission is Local AI and Infinite Intelligence. It is guided by an uncompromising vision of a world where AI lives and learns on any device, empowering individuals and communities rather than concentrating power in corporate data centers.

About Tether Data

Tether Data, S.A. de C.V. (“Tether Data”) is part of Tether’s broader vision to advance freedom, transparency, and innovation through technology. Its mission is to enable people and organizations to connect and share information directly, without unnecessary intermediaries. By creating secure, peer-to-peer systems, Tether Data gives users greater control over their data, communications, and digital interactions. Tether Data aims to redefine how information flows across networks by replacing centralized models with decentralized infrastructure designed for privacy, efficiency, and resilience. The company’s goal is to make global connectivity faster, safer, and more private, empowering individuals and institutions alike to exchange information freely and securely.