Hardware
Local AI needs real hardware numbers.
We will not label a PC “compatible” from a marketing guess. The public checker will use validated AVRE benchmark profiles.
Hardware checker preview
LLM
Model size, context and quantisation affect memory use.
Voice
STT/TTS add their own CPU/GPU and memory load.
Images
Image generation can be one of the heaviest local GPU workloads.
3D avatar
The renderer must share GPU resources rather than assuming it owns the machine.