Relay, founded by two former Nothing employees, is launching an AI-powered voice-to-text app for macOS now and plans a dedicated hardware microphone, the Relay Q, in early 2027. The system transcribes, performs contextual actions, and routes data through Google’s Gemini models while keeping local copies on your device.
Relay provides a macOS-first voice-to-text app that activates with a hotkey and uses Google’s Gemini models to transcribe and take contextual actions.
Relay can automate tasks (called Skills) like drafting messages or creating calendar events, but those automations require broad system permissions and data flows that include Relay’s servers and Google’s processing.
# What Relay is building Relay launched a macOS voice-to-text app now and plans to ship a dedicated microphone, the Relay Q, in early 2027. The company was founded by Cookie Xu and Raymond Zhu, both previously at Nothing. Relay's premise: voice can be a faster, more natural way to get ideas into text—especially at a desktop, where conversations and tone are more complex than on phones.
# How it works on macOS today During setup you assign a hotkey (default: Fn). Hold the hotkey while your cursor sits in any text field and Relay captures your voice, transcribes it with Google's Gemini models, and inserts text. The app can also run contextual automation called Skills—examples include drafting a Slack message, creating a Google Calendar event, or composing email.
# The Relay Q hardware concept Relay describes the Relay Q as a small cylindrical microphone intended to capture high-fidelity, low-volume speech. The hardware is meant to solve a common office problem: whispering so only your device hears you. Zhu tells WIRED that at Nothing they used Wispr Flow for transcription but struggled to get the right volume in open offices, which motivated building dedicated hardware.
# Privacy and data flow
# Beta behavior and accuracy
# Automation and friction points
# How Relay fits into the landscape Relay joins other voice-first tools and dedicated hardware efforts aimed at reducing typing friction. Competing approaches include on-device dictation advances (Google's Pixel features), standalone voice buttons with their own mics, and phone-as-mic workarounds. Relay's differentiator is an integrated app-plus-hardware strategy focused on desktop productivity and contextual automations.
# Practical considerations for readers If you're frustrated by typing and want to test voice-first desktop workflows, Relay's macOS app is an option to try in beta. Expect imperfect transcription, upfront permission prompts, and varying automation reliability. If you work in shared office spaces, the Relay Q's whisper-capture aim may address privacy and noise concerns, but that hardware isn't shipping until early 2027.
This voice-to-text mic, which is also wearable, is a stylish AI peripheral from former Nothing designers.
Voice input on phones has been solved for years. What has not been solved is the output. Speak into most dictation tools and you get back exactly what you said, fillers and false starts included, in a note you then have to clean up and move somewhere else. SpeakON attacks that gap with hardware: a […] The post SpeakON
I write for a living and honestly, the process of typing is my bottleneck. Whether it’s getting my braindumps down on paper, sending email replies, or just simply writing out an article, I can verbally communicate my ideas faster than I can type them. That’s why I love dictation tools. And I think anyone can benefit fr
Imagine a workplace where you spend less time attached to a keyboard, worrying about how to word an email , write a report, craft a LinkedIn post, polish a presentation deck, or sound professional enough in an office-wide memo. Instead, you stop hunching over a keyboard and start talking your way through a written corr
The Ulanzi AU05 VibeKey AI Voice Input Keypad is a bit of an unusual product. It is being touted as a compact AI workflow controller for photo and video creators. Ulanazi recommends it for use with mainstream AI voice-input software. It is also compatible with AI agents such as ChatGPT, Claude, and Codex, plus compatib
Before spending on a better mic, give that thing in your pocket another chance.
Loading more related stories...
Open the app view to save this story, compare related coverage, and continue from the same source.