Relay, a London startup founded by two former Nothing employees, launched its macOS dictation app today and previewed a $150 dedicated microphone called the Relay Q that will ship in the first quarter of 2027. The software runs on Google Gemini models and is bundled with one year of the subscription, which costs $15 per month afterward. Buyers do not need the hardware to use the app.
The company was started by Cookie Xu, Nothing's former head of AI and services, and Raymond Zhu, a hardware product manager at the same company. Both said they were heavy users of Wispr Flow's dictation software while at Nothing, and the friction of talking to a laptop microphone in a shared office is what pushed them toward building dedicated hardware. Relay is macOS-only at launch, with Windows, iOS, and Android on the roadmap.
Zhu says the desktop is the harder problem than mobile. Personal messages on a phone are simple, but work conversations at a computer involve brainstorming, varied tone, and shifting formality — the kind of speech that exposes the limits of a laptop's built-in mic.
Key facts
- 01Relay Q hardware will cost $150 and include one year of software; the subscription is $15 per month after that.
- 02The macOS app ships today; the Relay Q microphone and mobile apps arrive in Q1 2027.
- 03Relay was founded by Cookie Xu, former head of AI and services at Nothing, and Raymond Zhu, a former Nothing hardware PM.
- 04Voice-to-text is powered by Google Gemini models, with Skills that trigger actions in Slack, WhatsApp, Gmail, and Google Calendar.
- 05Competing hardware includes Sandbar's Stream Ring and Pebble's Index 01, both starting to ship to initial buyers.
The macOS app binds to a hotkey — Fn by default — that activates dictation inside any text field. Beyond transcription, Relay supports contextual actions it calls Skills: telling it to message a colleague in Slack prompts the app to open Slack, find the recipient, and draft the text. Presets exist for WhatsApp, Gmail, Slack, and Google Calendar, and users can wire up their own.
Those Skills require broad system permissions, including microphone access and, optionally, permission to capture a screen snapshot each time Relay activates. Relay says data is stored locally on the device and passes through its servers only in transit to Google for processing, with no retention on Relay's side. Handing an AI model that level of visibility into an operating system still carries the same category of privacy trade-off seen with Microsoft Recall.
The hardware itself is a cylindrical device with a microphone on one end that looks like a letter Q from above. Users hold down the top to dictate and double-tap to send. The button detaches from the cylinder and clips to a shirt if the user does not want to hold it, and Zhu says whispering directly into it should be enough for accurate capture — the specific pain point that Wispr Flow on a laptop could not solve.
Relay considered a ring form factor but decided against it, citing a crowded market. Sandbar unveiled the Stream Ring last year and Pebble followed with the Index 01, and both are now starting to ship to initial buyers. Relay Q ships with a charging dock and an optional MagSafe grip that attaches to an iPhone or Qi2-enabled device, letting the same button trigger dictation on mobile when the phone apps launch alongside the hardware in Q1 2027.
Transcription quality is early. The beta app does not capture every word perfectly, punctuation is inconsistent, and emoji insertion is hit or miss — asking for a heart emoji sometimes produced the emoji and sometimes produced the literal words "heart emoji." Google's Rambler feature on the Pixel 11 series currently sets a higher bar for accuracy. These are the sorts of rough edges that get sanded down in shipping updates, and dedicated hardware should shorten the noticeable delay between pressing the hotkey and activation.
Customization is where Relay is trying to differentiate. The app detects which contact you are writing to and can adjust tone accordingly — professional to a manager, casual to a friend, translated automatically to a coworker in another country. Users can add persistent notes, such as instructing it to skip periods in Slack messages, and the system honors those rules.
The category is getting crowded quickly. Wispr Flow already handles meeting transcription and cleanup, Google is pushing on-device dictation on Pixel, and hardware entrants like Sandbar and Pebble are shipping wearable microphones aimed at the same whisper-to-write use case. Relay's bet is that a purpose-built handheld with an app-level Skills layer beats a general-purpose ring or a built-in laptop mic on both accuracy and workflow depth.
Voice as the primary input to a computer has been promised for two decades and mostly failed on accuracy and latency. What changed is that large language models are now good enough to clean unscripted speech into usable prose, and the remaining bottleneck really is the microphone. If Relay ships hardware that captures a whisper cleanly and pipes it into an agentic layer that can actually complete tasks, the $150 price point is aggressive against wearable competitors and cheap against the software subscription's lifetime value. The harder question is whether granting one more startup screen-capture permissions on a work laptop will clear enterprise IT — and that will decide whether Relay Q lands as a consumer curiosity or a genuine keyboard replacement.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
