77 fully local images in 15s on an iPhone 13 by moving SDXS to the Neural Engine

QVAC_Official · reddit · 2026-11-19

The QVAC team made fully local image generation 20x faster on an iPhone 13 (2021, 4GB RAM): same SDXS model (512x512), same 15 seconds, airplane mode.

The only change was where the model runs: the UNet and tiny decoder were moved to the Neural Engine via Core ML inside their fork of stable-diffusion.cpp, keeping the text encoder on GPU. No server, no network.

Open source and free: tetherto/qvac-ext-stable-diffusion.cpp PR #66.

Original post →

More from Infra

Infra channel →