You need the software, but there’s nothing about that request that should require access to the Internet.

I have a LLM chatbot that controls my Home Assistant and Kodi players. It’s all done locally and the response time is under a second.

On my PC(Arch, btw) I have a global hotkey so I can hold the key to record a message and when I let go of the key it uses a local model to do speech to text and sends the result to the chatbot.

I could probably use a wake word but I’d need to mic up my house and I’d rather not do that. A bluetooth lapel mic and a single button Bluetooth “keyboard” about the size of a key switch (using an ESP32C3 microcontroller) give me the same functionality.