Menu

Jetson AGX Thor application demo: local conversation, camera input and online retrieval
📰

Jetson AGX Thor application demo: local conversation, camera input and online retrieval

#nvidia#redditnvidia#jetson#inference#delay
Reading 0:00
15s threshold

I’m developing Evopien on Jetson AGX Thor. Here’s a short application-level demonstration rather than a standalone inference benchmark.

The recording moves between voice conversation, a visual-monitoring request, online questions, multi language dialogue and a description of the live camera view. It ends with an interruption by voice.

The distinction I want to keep clear is local inference versus external information: the AI runs on the device, but weather and football queries use web retrieval.

My next engineering focus is sustained behavior under combined audio/video load, not just whether each component works separately. For people developing similar Jetson applications, what traces do you capture to distinguish input delay, inference delay and playback delay?

v.redd.it