Relay Q AI Microphone: Voice-to-Text Hardware for Builders

Relay, a London-based startup founded by former Nothing executives Cookie Xu and Raymond Zhu, is building a dedicated microphone called the Relay Q that pairs with macOS software to deliver high-fidelity voice-to-text powered by Google's Gemini models. The hardware won't ship until early 2027, but the software is available now in beta. For builders evaluating voice input workflows, the Relay Q AI microphone represents a hardware-software integration approach that aims to solve the office noise problem that software-only solutions struggle with.

A dedicated microphone for voice-to-text, powered by Google Gemini

The Relay Q is a cylindrical microphone with a button on top. You press and hold it to trigger voice input on your computer or phone, or you can clip the button to your shirt for hands-free use. The device comes with a charging dock and a MagSafe grip for iPhone. On the software side, the macOS app activates via a hotkey (default Fn) in any text field. The voice-to-text engine runs on Google's Gemini models.

Relay also offers contextual actions called Skills. You can ask it to message a colleague in Slack, create a Google Calendar event, draft a Gmail message, or send a WhatsApp text. The app can detect who you are talking to and adjust tone, formality, or even translate into another language. You can create custom Skills for other apps.

What Relay Q means for voice input workflows

For builders shipping AI products or internal tools, Relay Q is interesting because it tackles a specific pain point: reliable voice input in an open office. The founders experienced this firsthand at Nothing, where they used Wispr Flow but struggled with volume control. Dedicated hardware lets you whisper quietly and still get accurate transcription, which software-only solutions often fail at.

The Skills system also points toward a future where voice commands can trigger multi-step workflows across apps. If you are building a voice-enabled assistant or agent, Relay Q's approach of combining a hardware trigger with app-level automation is worth watching. The company plans to expand to Windows, iOS, and Android later, which could make it a cross-platform input layer.

Privacy, permissions, and the trade-offs

Relay Q requires broad system permissions. It needs microphone access and can record a snapshot of your screen when activated to execute Skills. The company says data is stored locally on your device, where you can edit or delete it. When sent for processing, it passes through Relay's servers to Google's servers. Relay claims it does not retain the data. Still, granting an AI model screen access introduces privacy trade-offs similar to Microsoft's Recall. Builders evaluating this for enterprise use should consider data residency and compliance requirements.

Caveats and open questions

In WIRED's beta testing, the software's transcription accuracy lagged behind Google's on-device Rambler feature on Pixel phones. Punctuation was sometimes off, and emoji insertion was inconsistent. The hardware may improve latency and accuracy, but that won't be clear until 2027. The software is macOS-only at launch, and the $150 hardware price plus $15/month subscription adds up over time. Competitors like Sandbar's Stream Ring and Pebble's Index 01 are already shipping, though they take a ring form factor. Relay Q's value proposition depends on whether dedicated hardware meaningfully outperforms software-only alternatives in real office conditions.

FAQs

Relay Q is a dedicated microphone and macOS software for high-fidelity voice-to-text powered by Google's Gemini models. It activates via a hotkey or a button on the microphone and can perform contextual actions (Skills) in supported apps like Slack, Gmail, and Google Calendar. Processing involves local data storage, Relay servers, and Google's servers; Relay says it does not retain user data.