Wispr Flow is tackling the perennial tech problem, failed voice assistants. While companies like Apple and Amazon have struggled to move Siri and Alexa beyond simple timers and weather reports, Wispr is taking a radically different approach.
Wispr’s primary thesis is that voice AI doesn’t work because they don’t build sticky habits. Instead of trying to be a “do-everything” AI app, Wispr is only focusing on one high frequency action, voice dictation. Wispr Flow aims to replace typing. Their data shows that after six months, power users rely on their keyboards for only 28% of their computer input. By making voice dictation work on the first try, they are challenging the assumption that voice AI tech simply doesn’t work.
Once Wispr Flow has mastered voice dictation, it plans for deep integration with user’s everyday workflows. They are avoiding flashy but infrequent tasks such as booking a flight, in favour of tasks that people do everyday. Background messaging – allowing users to send work messages on platforms such as Slack, without opening the app, bypassing the distraction spiral of unread notifications. Contextual Assistance – enabling users to ask questions about what is currently on their screen or get explanations for cryptic messages without context switching. “No more AI Slop” – Wispr Flow is moving away from the traditional copy-paste workflow on ChatGPT, Wispr wants to polish and edit text directly within the user’s flow, personalizing to their unique voice and style. Communication coaching – Wispr Flow aims to become a communications coach for users, making them more effective and life and at work.
Wispr views current computer interfaces as distracting and sources of strain. Their ultimate goal is to move software to the background. Allowing users to communicate their intent via voice without having to be precise. They aim to make technology as simple as communicating with a friend.
So I gave Wispr Flow a try and here are the results. You can use Wispr Flow for drafting messages, emails, documents and even annotating patient notes. The app gave me an interface that was connected to Slack, allowing me to send messages via voice. The app was pretty accurate though it made some mistakes, but it was still pretty cool.
Next, I used Wispr Flow to create an email via voice, deliberately trying to not be precise and they AI faired fairly ok. It made some mistakes but it did get the context of my message and was able to indent various paragraphs.
If you are intrigued, give Wispr Flow a try over here.