Running SD Natively on Android: SDXL Takes 20 Minutes on a Phone
Silent-Paramedic4063 · reddit · 2026-08-09
A developer built StableDroidfusion entirely from scratch using Termux, successfully running Stable Diffusion models natively on Android. Utilizing the Vulkan GPU pipeline on a Pixel 9 Pro (16GB RAM), the tool can handle SD1.5 and SDXL/Pony models.
For performance, generating a 512x512 image with SD1.5 LCM takes about 2 minutes, while a 1024x1024 image with SDXL requires around 20 minutes. The initial GPU pipeline build takes 30 minutes but is subsequently cached, and the model can be kept warm in VRAM to reduce load times for subsequent generations. The project is still a work in progress, with plans to open-source it in the future.
More from Infra
- MLCommons launches MLPerf Endpoints, a benchmark that measures inference endpoints the way buyers actually purchase them — TheKanter · 2026-10-16
- Neocloud M&A decoded: CoreWeave, Nebius, Crusoe, Nscale all bet margins vanish beyond GPU scarcity — AccBalanced · 2026-10-16
- AI-Driven Timing-Aware Power Optimization Uses Learned Ranking With Baseline-Preserving Repair — Amit Sharma · 2026-10-16
- Nscale AI datacenter plagued by coolant leaks, transformer fire, 10 safety violations — saranormous · 2026-10-16
- WSL 3.0 ships with native containers and Linux 6.18 kernel, syscall overhead down ~16% — solyarisoftware · 2026-10-16
- Are 22 GB modded RTX 2080 Ti cards a viable budget local LLM option? — Dave86ch · 2026-10-16