speak to type
weekly startup spotlight: wispr flow 🎙️
most people speak at 150+ words per minute, but type at only 40-60 words per minute (my monkeytype score begs to differ…). existing dictation tools are often inaccurate, fail to understand context, and require manual editing to remove filler words or fix formatting - the additional effort to correct isn’t worth it for many.
wispr flow, founded in 2021, is an ai-powered dictation application that runs at the system level to turn natural/messy speech into perfectly formatted text in any app. it automatically removes filler words (”um” and “like”), corrects mid-sentence backtrack (ex. “meet at 2, no make it 3”), and adapts its tone based on whether a user is on slack, email, or even a code editor. the company uses a multi-layered ai architecture that first transcribes audio, and then passes it through llms to handle real-time editing and context-aware formatting. users just press hotkeys (like fn on mac) to “start recording” audio and then release to paste the text onto an app interface.
to date, wispr flow has raised $81 million in funding. the company also reports a 40% month-over-month growth rate, with its tool used across 25,000+ apps.
questions i’d want to ask:
how do you distinguish between intentional dictation and passive background audio to prevent accidental data capture?
how do you ensure that moving from typing to speaking doesn’t make users “the loud person” in quiet offices or cafes?
today’s drops 🔍️
3 weeks left to participate in the $40,000 amazon nova ai hackathon
check out this extensive internship list