FPS: --
Frame: 0
Initializing...
Diagnostics will appear here once inference starts.
Display Speed
Display frames as fast as the worker produces them
Gate display to 25 FPS, synced to audio clock during streaming
Inference Backend (requires reload)
Try GPU first, fall back to CPU if unavailable
Fail if WebGPU is unavailable
Use CPU — try this if WebGPU is slow on integrated GPUs
GPU Preference (requires reload)
Default — targets typical user hardware
Testing only — use dedicated GPU (e.g. RTX/Radeon) if available
Benchmark
Runs a quick benchmark on each available backend and logs results to Diagnostics