It's also a desktop app, integrated into the mobile app, and it has cloud hosted session capabilities too. Plugs into Slack and Github, maybe others. So definitely there's backend stuff as well, beyond the inference.
I picked up a couple of foot pedals built to control manual transcription software from thrift stores. That was years ago and I hacked them to control music software instead. Then one day I got Nvidia's Parakeet model running locally and was sold on automatic transcription in an instant. So in an ironic twist, I started using one of my spare transcription pedals to do transcription by wiring it up to a harness for Parakeet. It's great and the setup is basically trivial to create these days on Linux with a bit of help from an agent.