Tools for speaking, working with AI, describing computation and making videos.
A macOS input source with personal vocabulary, designed to fit the apps you already use. Speech recognition runs on the device.
avette connects our speech research to an everyday tool. The wider work includes a player for transcription, iPhone input and live conversations.
A workspace for small companies and schools. Rooms hold the conversation, the work and its history. Agents can use the machines you give them access to, with your choice of models and inference providers.
Built with Aurora Nockert. European jurisdiction is a design priority, alongside control over where computation happens.
vix describes computations and their dependencies, so work can be reused when inputs have not changed. It is a language and a runtime, not just another build command.
Build systems, distributed builds and fast CI are uses of the same underlying idea. The project is open source; vix Cloud is planned as a managed service.
Our replacement for DaVinci Resolve, tailored to the way fasterthanlime videos are made. It is in development for our own use.