BUD-E — short for Buddy for Understanding and Digital Empathy — is an open-source conversational voice assistant developed by LAION together with the ELLIS Institute Tübingen, Collabora and the Tübingen AI Center. It aims for real-time, natural spoken conversation with emotional awareness, running fully locally on consumer hardware. The GitHub repository has around 500 stars and was last updated in 2024.
Most voice assistants are cloud services with stilted turn-taking. BUD-E is a research system exploring a different point in the design space: a fully local assistant that replies in real time with natural voices, tracks long-term context from previous conversations, and handles the messy reality of speech — multiple speakers, interruptions, affirmations and thinking pauses. An empathy-focused design tries to read and respond to the user’s emotional state rather than treating speech as plain text commands. The stack combines open speech recognition, a language model and neural text-to-speech, orchestrated to keep latency low enough for fluid dialogue on a home machine, with no audio leaving the device. As a research project it doubles as a testbed for studying empathic spoken interaction.
The system is modular, so researchers can swap the speech recogniser, dialogue model or synthesiser to study each component’s effect on perceived empathy and naturalness. Its consortium backing — spanning LAION’s open-data community and Tübingen’s research institutes — reflects an explicitly open-science approach: code and findings are shared publicly rather than kept behind a product wall.
BUD-E is free and open source under the MIT license. There is no hosted service or paid tier; you run it on your own machine.
Researchers, hobbyists and privacy-conscious builders experimenting with local, empathic voice assistants and real-time spoken dialogue systems.