Talking to a chatbot feels different from typing — more fluid, more human. Increasingly, researchers argue that this shift isn’t just convenient but transformative. Voice changes how people think and interact with AI, pulling it out of the “search box” mindset and into something more collaborative. “If you haven’t tried voice chatting with an AI model to see the appeal, you should,” says University of Pennsylvania professor Ethan Mollick. “Anthropomorphism is the future, in ways good and bad.”
That future has now arrived in the world of R programming and the tidyverse. At last month’s posit::conf(2025) data science conference, a new AI assistant called ggbot2 was introduced — a voice-based chatbot designed to help users create ggplot2 visualizations simply by speaking.
Posit Chief Scientist Hadley Wickham demonstrated the system during his keynote, showing how a user could describe a plot out loud and watch ggbot2 generate both the visualization and the corresponding R code in real time. “My goal has always been for the code to get out of the way,” Wickham said. “I want people to express their ideas and interact with data as quickly as possible.”
A week after the conference, Posit officially released the ggbot2 package, available for anyone to install and try. Built using the shinyrealtime library — which connects R and Python Shiny apps with OpenAI’s Realtime API — ggbot2 enables low-latency, spoken conversations that feel almost natural. Users can ask for plots, tweak parameters, or request new visuals mid-conversation, and the bot instantly updates both the code and chart.
Behind the scenes, OpenAI’s Realtime API handles the voice streaming and conversation management, ensuring fast responses and support for user interruptions. According to Posit CTO Joe Cheng, this API was ideal for creating fluid, voice-driven data interactions. To experiment with ggbot2 yourself, you’ll need an OpenAI API key — and a willingness to talk to your code instead of typing it.