We made a fully animatronic GLaDOS! Inside the body is a Z1 from Unitree to move it around. It is always looking at the nearest person by using person-tracking and depth perception from a Stereolabs Zed2i. You can also talk to GLaDOS by using automatic speech recognition through RIVA. That will take audio and convert it into text. The text is passed as a prompt to a large language model that is running locally on the Jetson (Forge from ConnectTech). The response from the LLM is put through a custom text-to-speech pipeline (Fastpitch + Hifigan) to create the voice so she can talk back to you!
This is wild! Incredible work integrating all the bits and pieces together @davesarmoury! Even the 3D-printing and ASR/TTS/LLM are impressive on their own. Hope GLaDOS is a useful assistant for you in your lab :)