Wispr Advanced Interfaces Lab is the research group 📝Wispr Flow launched in July 2026 to build voice-first systems that infer user intent and generate whichever interface best serves it.
The lab was announced on July 24, 2026 by Ariya Rastrow, who joined Wispr as Chief Science Officer to lead it alongside co-founders Tanay Kothari and Sahaj Garg. Rastrow began work on speech recognition during a PhD at Johns Hopkins, joined 📝Amazon in 2012 as a founding member of the team behind 📝Alexa and Echo — eventually leading roughly 400 researchers and engineers — and later led next-generation voice and multimodal foundation model work for wearables and smart glasses at 📝Meta. The lab sits with Wispr's machine learning team in 📝San Francisco, CA.
Its stated premise is that foundation models have largely settled the intelligence question, leaving the interface as the bottleneck. Rastrow's summary — "AI is an interface problem now" — frames a decade of voice assistants whose top uses never moved past weather, music, timers, reminders, and news. The lab rejects the conventional speech-to-text-to-intelligence-to-text-to-speech pipeline in favor of what it calls voice to UI rather than voice to voice: interpreting speech within the context of a user's digital world, including memory, active application, open document, and current task, then producing whatever output fits — a spoken reply, text, an executed action, a full workflow, or a dynamically generated interface.
Structurally, Wispr positions itself as model-agnostic, operating an intelligence layer above the model stack that routes each request to the best backend model, tool, or service for quality, latency, cost, and experience, and building upward from 🏷️#speech-to-text dictation rather than from voice-in, voice-out demos. The lab's first public output is Canto, a real-time dictation speech model introduced in September 2026, which Wispr reports achieved the lowest word error rate among the models it evaluated on real-world dictations. The work extends the company's broader argument about 📝Human-Computer Interaction (HCI): reducing cognitive friction so intent translates directly into outcomes.